Shadab is available for hire
Hire ShadabShadab Arif Ansari
Verified Expert in Engineering
ML Engineer and Developer
Bengaluru, Karnataka, India
Toptal member since October 9, 2025
Bio
Shadab is a lead AI engineer with 8+ years of experience building production AI systems for healthcare and fintech enterprises. He specializes in agentic AI, multi-agent orchestration (LangGraph), and advanced RAG/GraphRAG architectures powered by Claude, OpenAI, and Llama. Shadab has deployed secure, high-throughput ML platforms on AWS and Azure—from real-time lending decisioning at scale to enterprise member-risk prediction. He's an IIT Kharagpur alumnus and a top 3 LLM engineer at Turing.
Portfolio
UnitedHealth Group
Agentic AI Systems, Agentic RAG Systems, Agentic Coding, Agentic Frameworks...
RVIN
Agentic AI, ChatGPT, Large Language Models (LLMs), Python...
Digital Exchange
Retrieval-augmented Generation (RAG), Agentic AI, Agentic RAG Systems...
Experience
- Python - 9 years
- FastAPI - 6 years
- LangChain - 4 years
- Retrieval-augmented Generation (RAG) - 4 years
- Generative Artificial Intelligence (GenAI) - 4 years
- LangGraph - 4 years
- Chatbots - 4 years
- Agentic AI - 3 years
Preferred Environment
Amazon Web Services (AWS), Azure, LangGraph, LangChain, PyTorch, Python 3, Pinecone, FAISS, Agentic RAG Systems, Agentic AI
The most amazing...
...: I shipped a multi-agent customer support RAG platform end-to-end—back-end, AI, and front-end—that served 10,000+ customers and cut support workload by 44%.
Work Experience
Lead AI Engineer
2026 - 2026
UnitedHealth Group
- Developed a GenAI-powered agent to automate claim intake, policy verification, document review, fraud detection, and claim summarization, improving processing speed and operational efficiency.
- Built a RAG-based AI agent that answers policyholder queries, explains coverage, exclusions, premiums, and claim eligibility using natural language, improving customer support efficiency.
- Created an LLM-based document processing solution to extract, classify, summarize, and query information from clinical notes, discharge summaries, and insurance documents using RAG and vector search.
- Built an ML-based member risk prediction model to identify high-risk members using healthcare and behavioral data, enabling proactive interventions, improved risk stratification, and data-driven care management.
Technologies: Agentic AI Systems, Agentic RAG Systems, Agentic Coding, Agentic Frameworks, Python, Machine Learning, Azure, Agentic Workflow Design, HIPAA, HIPAA Compliance, Amazon EKS, Healthcare, Whisper, RingCentral, Qwen, AI platform engineer, Machine Learning Infrastructure Engineer, Demand Forecasting, Model Deployment, Datasets, LLM Fine-tuning, Documentation, LLM Reasoning, Image Generation, Rasa NLU, Multimodal GenAI, Rasa.ai, Small Language Models (SLMs), Technical Architecture, GitLab CI/CD, Gemini, LoRa, QLoRA, BERT, SHAP, EDA, Matplotlib, Seaborn, JSON, Replit, Data Quality, Delta Lake, Apollo, Clay, Text to Image, Text to Image AI, Software Architecture, Technical Leadership, OpenAPI, Cortex, Back-end, Data Extraction, Google Document AI, Handwriting Recognition, Applied AI, Cloud Run, Product Engineering, SaaS, Startups, Combinatorial Optimization, Cost Modeling, Artificial Intelligence (AI), LLM Agents, OpenAI Agents SDK
Lead AI Engineer
2025 - 2026
RVIN
- Led development of a multi-agent RAG system using LangGraph, Azure OpenAI, Azure AI Speech, AKS, and Azure AI Search for Arabic voice/text support, reducing support tickets by 44% and improving customer resolution rates by 28% overall.
- Built an autonomous LangGraph scheduling agent with MCP tool integrations for booking, rescheduling, and cancellations—integrated with Databricks-backed data workflows and deployed on Azure and Kubernetes for scalable, zero-touch workflow automation.
- Developed a GraphRAG platform using Neo4j, LangGraph, and vector search to connect policies, SOPs, and enterprise documents, enabling multi-hop reasoning and contextual AI search across large-scale enterprise knowledge bases.
- Built an LLM pipeline to extract organizations, job roles, locations, and relationships from unstructured documents and webpages using structured prompting, validation, and evaluation workflows for high-accuracy data generation.
- Created an AI-powered email reply agent using Microsoft Copilot that reads your inbox, understands context, and auto-drafts smart responses — boosting productivity with LLMs and Gmail API.
- Developed a multimodal AI chatbot with Llama, Claude APIs, Whisper, TTS, image generation, vector memory, intent detection, and context-aware personalized responses.
- Built a personalized AI companion using LLMs, RAG, Pinecone, Rasa, voice AI, semantic memory, prompt caching, RL-based personalization, and scalable data pipelines.
Technologies: Agentic AI, ChatGPT, Large Language Models (LLMs), Python, Retrieval-augmented Generation (RAG), OpenAI API, Meta Llama, Artificial Intelligence (AI), Web Scraping, Vector Databases, ChatGPT API, Document Processing, Large Language Model Operations (LLMOps), Document Parsing, Docker, CI/CD Pipelines, Pydantic, Redis, Kubernetes, Amazon EC2, Data Engineering, Llama 3, Tesseract, Azure AI Document Intelligence, Chatbot Conversation Design, Open-source LLMs, Speech-to-Text (STT), Text-to-Speech (TTS), RAG Pipelines, AI Voice Agents, GitHub, Google Calendar, AI-generated Video, API Development, Semantic Search, Data Governance, AI Architecture, Beautiful Soup, Website Data Scraping, RAG Systems, Custom Models, OpenAI GPT-4 API, Azure AI Search, Azure OpenAI Service, RESTFul APIs, Foundry, Microsoft Copilot Studio, Azure AI Studio, Microsoft Power Platform, Low Code, AutoGen, Semantic Kernel (SK), Claude Code, Claude, Financial Analysis, Reporting, Observability, Observability Tools, Observable Framework, AI Pipeline, Grafana, AI Agents, CRM APIs, Debugging, Workflow Automation & System Integration, AI SDK, API Design, Platform Design, Scraping, WhatsApp API, WhatsApp, Chatbots, Generative Artificial Intelligence (GenAI), LangGraph, LangChain, AWS IoT, Scalable Vector Databases, LangSmith, Amazon ElastiCache, Model Context Protocol (MCP), Benchmarking, Asynchronous Programming, Cloud Platforms, Event-driven Architecture, Multi-agent Systems, Model Evaluation, System Architecture, Webhooks, Claude Agent SDK, Supabase, n8n, Twilio, ElevenLabs Solutions, Qdrant, ColBERT, Cloud, Product Development, React, Vercel, LlamaIndex, Kubeflow, Neo4j, Graph Databases, SAS, AI Agents, Agentic AI, Artificial Intelligence (AI), Generative Artificial Intelligence (GenAI), Large Language Models (LLMs), Python, Product Management, Amazon Web Services (AWS), Product Development, Asyncio, Azure Kubernetes Service (AKS), Event-driven Design (EDD), System Design, OpenAI SDK, Platform Engineering, Snowflake, Hugging Face Transformers, Visual Language Models (VLMs), Ollama, Technical Writing, AI Marketing, Artificial Intelligence as a Service (AIaaS), Data Lakes, Marketing, Coding, Leadership, Team Leadership, SDKs, Vector Search, Unity Catalog, Agentic AI Systems, Business Intelligence (BI), Governance, Regression Testing, Cloud Architecture, Data Architecture, Enterprise Architecture, Solution Architecture, Stakeholder Management, Decision Modeling, Data Warehousing, Embedding Models, JavaScript, Django, Azure AI Services, AI Agent Orchestration, Harness, Light LLMs, Pgvector, UiPath, DevOps, Streamlit, Azure Cosmos DB, Multiagent Generative Systems (MAGs), GraphRAG, LLM Integration, Semantic Code, Azure Databricks, Langfuse, Azure Machine Learning, Azure ML Studio, Cloud Governance, Palantir Foundry, API Gateways, Agentic Coding, Audio Processing, Automatic Speech Recognition (ASR), Distributed Systems, Telemetry, Monitoring, Cursor AI, Industrials, Azure Functions, Azure Function App, Amazon Bedrock AgentCore, Java, Full-stack, MongoDB, Application State Management, Next.js, Node.js, Codex, Data Security, Quantization, Git, Datadog, AI Hallucinations Management, Agentic Workflow Design, Whisper, RingCentral, Qwen, Machine Learning Infrastructure Engineer, Demand Forecasting, Model Deployment, Datasets, LLM Fine-tuning, Documentation, LLM Reasoning, Image Generation, Rasa NLU, Multimodal GenAI, Rasa.ai, Small Language Models (SLMs), Technical Architecture, GitLab CI/CD, Gemini, LoRa, QLoRA, BERT, EDA, Matplotlib, Seaborn, JSON, Replit, Data Quality, Delta Lake, Text to Image, Text to Image AI, Software Architecture, Technical Leadership, OpenAPI, Cortex, Back-end, Data Extraction, Google Document AI, Handwriting Recognition, Applied AI, Cloud Run, Product Engineering, Early-stage Startups, SaaS, Startups, Combinatorial Optimization, LLM Agents, OpenAI Agents SDK
Senior AI Engineer
2025 - 2025
Digital Exchange
- Designed and deployed a LangGraph and Claude AI agent for patient intake, query resolution, appointment workflows, and medical form extraction — delivered end-to-end independently for clinical use.
- Built and deployed an AI assistant to analyze lab reports and clinical documents, enabling healthcare teams to retrieve insights and summaries instantly — shipped independently for a live client.
- Created a healthcare support AI agent using LangGraph and Claude to answer patient queries, retrieve treatment information, and assist appointment workflows—designed, built, and delivered the full solution independently.
- Built a personalized recommendation engine using user behavior and transaction data; implemented feature pipelines, affinity scoring, and real-time inference, achieving ~80% AUC with strong precision and recall, improving recommendation relevance.
Technologies: Retrieval-augmented Generation (RAG), Agentic AI, Agentic RAG Systems, AI Architecture, AWS IoT, Scalable Vector Databases, Artificial Intelligence (AI), API Design, Platform Design, Hyperparameter Tuning, Benchmarking, Asynchronous Programming, Cloud Platforms, Event-driven Architecture, System Architecture, Cloud, Product Development, Vercel, LlamaIndex, Graph Databases, AI Agents, Agentic AI, Artificial Intelligence (AI), Generative Artificial Intelligence (GenAI), Large Language Models (LLMs), Python, Product Management, Amazon Web Services (AWS), Azure, Product Development, Asyncio, Azure Kubernetes Service (AKS), System Design, OpenAI SDK, Platform Engineering, Claude Agent SDK, Snowflake, Hugging Face Transformers, Visual Language Models (VLMs), Ollama, Technical Writing, AI Marketing, Artificial Intelligence as a Service (AIaaS), Data Lakes, Coding, Leadership, Team Leadership, SDKs, Vector Search, Unity Catalog, Agentic AI Systems, Business Intelligence (BI), Governance, Regression Testing, Cloud Architecture, Data Architecture, Enterprise Architecture, Solution Architecture, Stakeholder Management, Bioinformatics, Clinical Research, Decision Modeling, Clustering Algorithms, Recommendation Systems, Prediction Markets, Imbalanced-learn, ONNX Runtime, Open Neural Network Exchange (ONNX), Go, Data Warehousing, Embedding Models, JavaScript, Django, Azure AI Services, AI Agent Orchestration, Harness, Light LLMs, Pgvector, UiPath, DevOps, Graphics Processing Unit (GPU), Bayesian Statistics, Streamlit, Azure Cosmos DB, LLM Integration, Semantic Code, Large-scale Data Processing, Cloud Governance, API Gateways, Agentic Coding, Audio Processing, Automatic Speech Recognition (ASR), Distributed Systems, Telemetry, Monitoring, Cursor AI, Industrials, Azure Function App, Amazon Bedrock AgentCore, MongoDB, Data Security, Quantization, Git, Datadog, AI Hallucinations Management, Agentic Workflow Design, Healthcare, Whisper, RingCentral, Machine Learning Infrastructure Engineer, Demand Forecasting, Model Deployment, Datasets, LLM Fine-tuning, Documentation, LLM Reasoning, Image Generation, Rasa NLU, Multimodal GenAI, Rasa.ai, Small Language Models (SLMs), Technical Architecture, GitLab CI/CD, Gemini, LoRa, QLoRA, BERT, EDA, Matplotlib, Seaborn, JSON, Data Quality, Delta Lake, Software Architecture, Technical Leadership, OpenAPI, Cortex, Back-end, Data Extraction, Google Document AI, Handwriting Recognition, Applied AI, Cloud Run, Product Engineering, Early-stage Startups, SaaS, Startups, Combinatorial Optimization, Cost Modeling, LLM Agents, OpenAI Agents SDK
Senior AI Engineer
2024 - 2025
An Online Freelance Agency
- Built a LangGraph + Azure OpenAI email agent integrated with Outlook via Microsoft Graph API; context-aware drafting at 3s latency, reducing manual email effort by 40% in production.
- Fine-tuned LLMs on domain-specific datasets using LoRA and QLoRA, improving contextual accuracy by 28% for compliance-critical applications across legal, medical, and finance verticals.
- Architected a LangGraph multi-agent chatbot with STT and context-aware recommendations across 10,000+ products — increased customer engagement by 35% in production.
- Delivered a LLaMA + RAG invoice agent, achieving 90% field extraction accuracy across 10,000+ invoices — reduced manual processing effort by 65% in live workflows.
- Designed a LangGraph + RAG claims agent that ingests policy docs and medical bills, verifies coverage, flags fraud risks, and autonomously drafts claim decisions end-to-end.
- Fine-tuned Qwen2.5-VL-3B with QLoRA (4-bit NF4 + LoRA r=16) on invoice receipt dataset for structured JSON extraction, achieving 90.2% strict / 97.8% normalized macro exact-match with 100% JSON parse rate on held-out test set.
Technologies: Python, Retrieval-augmented Generation (RAG), Agentic AI, Chatbots, Natural Language Processing (NLP), Machine Learning, REST APIs, OpenAI, Artificial Intelligence (AI), Data Science, Agentic Frameworks, Amazon Web Services (AWS), Azure, Databricks, Pandas, NumPy, Pinecone, APIs, Predictive Analytics, Automation, Data Pipelines, Data Collection, Web Scraping, Architecture, AI Model Integration, AI Modeling, AI Model Training, Vector Databases, eCommerce, Anthropic, PostgreSQL, Google Cloud Platform (GCP), AI Automation, ChatGPT API, AI Chatbots, FAISS, Document Processing, Multimodal Models, Document Parsing, Microservices, MLflow, Docker, Reinforcement Learning, Microsoft PowerPoint, Minimum Viable Product (MVP), Pydantic, Redis, Kubernetes, Amazon EC2, Data Engineering, Data Handling, Workflow, Deployment, Microsoft Copilot, Llama 3, Object Detection, Business Analysis, AI Consulting, AI Design, OAuth, Content, Generative Pre-trained Transformers (GPT), Amazon Textract, Tesseract, Azure AI Document Intelligence, Workflow Automation, RAG Pipelines, Reliability, Testing, AI Assistants, Agentic RAG Systems, RAG Architecture, Claude API, AI Voice Agents, Auth, ETL Pipelines, Model Validation, AI Copilots, OpenAI API, GitHub, Microsoft Excel, Data Privacy, API Development, Scalability, Semantic Search, Docker Swarm, Data Governance, Data Scientist, AI Architecture, RAG Systems, OpenAI GPT-4 API, Azure AI Search, Azure OpenAI Service, RESTFul APIs, GraphQL API, Foundry, Microsoft Copilot Studio, Azure AI Studio, Microsoft Power Platform, Low Code, CrewAI, TypeScript, Terraform, Financial Analysis, AI Pipeline, AI Agents, Debugging, Platform Design, Generative Artificial Intelligence (GenAI), LangGraph, LangChain, Consulting, AWS IoT, Scalable Vector Databases, Image Segmentation, Model Context Protocol (MCP), AI SDK, API Design, Benchmarking, Asynchronous Programming, Cloud Platforms, Event-driven Architecture, Multi-agent Systems, Model Evaluation, System Architecture, Cloud, Product Development, React, Vercel, LlamaIndex, Neo4j, Graph Databases, AI Agents, Agentic AI, Artificial Intelligence (AI), Generative Artificial Intelligence (GenAI), Large Language Models (LLMs), Python, Product Management, Amazon Web Services (AWS), Azure, Product Development, Asyncio, Azure Kubernetes Service (AKS), System Design, OpenAI SDK, Platform Engineering, Claude Agent SDK, Hugging Face Transformers, Visual Language Models (VLMs), Synthetic Data Generation, Ollama, Technical Writing, AI Marketing, Coding, Leadership, Team Leadership, SDKs, Finance, iGaming, Vector Search, Unity Catalog, Agentic AI Systems, Business Intelligence (BI), Governance, Regression Testing, Cloud Architecture, Data Architecture, Enterprise Architecture, Solution Architecture, Stakeholder Management, Decision Modeling, Data Warehousing, Embedding Models, JavaScript, Django, AI Agent Orchestration, Harness, Light LLMs, Pgvector, UiPath, DevOps, vLLM, NVIDIA TensorRT, Streamlit, Fine-tuning, LLM Integration, Semantic Code, Forecasting, Cloud Governance, API Gateways, Agentic Coding, Audio Processing, Automatic Speech Recognition (ASR), Distributed Systems, .NET, ASP.NET, C#, Telemetry, Monitoring, Cursor AI, Industrials, Amazon Bedrock AgentCore, Git, AI Hallucinations Management, Agentic Workflow Design, Machine Learning Infrastructure Engineer, Model Deployment, Model Development, Datasets, Documentation, LLM Reasoning, Image Generation, Rasa.ai, Small Language Models (SLMs), Technical Architecture, GitLab CI/CD, Gemini, LoRa, QLoRA, BERT, SHAP, EDA, Matplotlib, Seaborn, JSON, Replit, Data Quality, Delta Lake, Software Architecture, Technical Leadership, OpenAPI, Back-end, Data Extraction, Google Document AI, Applied AI, Cloud Run, Product Engineering, Early-stage Startups, SaaS, Startups, Combinatorial Optimization, Cost Modeling, LLM Agents
Senior Data Scientist
2022 - 2024
Paytm
- Built a financial document intelligence platform using hybrid RAG to extract and validate data from statements, PDFs, and spreadsheets with provenance tracking, audit trails, and deterministic calculations.
- Designed a customer engagement scoring machine learning model using app usage, recharge/payment frequency, transaction diversity, and retention behavior to improve campaign targeting and personalized recommendations.
- Built a fraud risk machine learning model with lightgbm, using transaction velocity, device signals, location patterns, and behavioral anomalies to identify suspicious merchants and payment activities in near real-time.
- Built stacked XGBoost and BERT-based models on GCP Vertex AI, leveraging over 35 million records to achieve a 75–76% AUC. Powered real-time risk alerts, loan underwriting, and portfolio intervention at Paytm scale.
- Architected comprehensive ML pipelines on SageMaker and EC2, incorporating CI/CD, MLflow tracking, Docker, and hyperparameter tuning. Established a production-grade training and deployment infrastructure for credit risk models.
- Built merchant clustering and segmentation models using transaction trends, business categories, geography, and growth signals to support targeted marketing, cross-sell opportunities, and business prioritization strategies.
Technologies: Python, PySpark, SQL, Machine Learning, Deep Learning, Generative Artificial Intelligence (GenAI), Apache Airflow, Model Monitoring, Feature Engineering, API Integration, REST APIs, OpenAI, Artificial Intelligence (AI), Data Science, Agentic Frameworks, Amazon Web Services (AWS), Databricks, Machine Learning Operations (MLOps), Pandas, NumPy, Data Analytics (Marketing), Marketing Analytics, Predictive Analytics, Marketing Mix Modeling, Data Pipelines, Data Collection, Predictive Maintenance, AI Modeling, Vector Databases, AI Automation, ChatGPT API, AI Tools, AI Chatbots, FAISS, Document Processing, Large Language Model Operations (LLMOps), Banking & Finance, Financial Systems, Document Parsing, Statistical Modeling, Spark, Time Series Forecasting, Credit Underwriting, Optuna, Predictive Modeling, Performance Marketing, Gradient Boosting, Time Series, K-means Clustering, Dimensionality Reduction, Pydantic, Data Engineering, Data Handling, Scientific Data Analysis, Workflow, RAG Pipelines, Project Management, GitHub, Microsoft Excel, Workflows, API Development, ML Pipelines, Data Governance, Data Scientist, Jupyter Notebook, Risk Modeling, Risk Models, ETL Tools, RAG Systems, OpenAI GPT-4 API, Scikit-learn, LightGBM, Hugging Face, Credit Risk, Financial Data, Fraud Detection, Financial Data Analytics, RESTFul APIs, Low Code, AI Pipeline, Debugging, Probabilistic Modeling, Vertex AI, Logistics & Supply Chain, Retrieval-augmented Generation (RAG), Chatbots, LangGraph, LangChain, AWS IoT, Image Segmentation, API Design, Platform Design, Hyperparameter Tuning, Benchmarking, Rule-based Programming, Asynchronous Programming, Cloud Platforms, Event-driven Architecture, Model Evaluation, System Architecture, Cloud, Product Development, Vercel, Kubeflow, Regression Modeling, Pricing Elasticity, Pricing Models, Statistical Analysis, Statistical Methods, Large Language Models (LLMs), Python, Machine Learning, Product Management, Amazon Web Services (AWS), Product Development, Asyncio, Coding, Finance, Business Intelligence (BI), Stakeholder Management, Decision Modeling, Clustering Algorithms, Prediction Markets, Imbalanced-learn, ONNX Runtime, Open Neural Network Exchange (ONNX), Data Warehousing, ETL Development, JavaScript, Django, High-performance Computing (HPC), DevOps, Graphics Processing Unit (GPU), Bayesian Statistics, Churn Analysis, Large-scale Data Processing, Random Forests, API Gateways, Monitoring, Amazon SageMaker, GIS, GeoPandas, Git, Machine Learning Infrastructure Engineer, Model Deployment, Model Development, Datasets, Documentation, Image Generation, Technical Architecture, GitLab CI/CD, BERT, SHAP, EDA, Matplotlib, Seaborn, JSON, Data Quality, Delta Lake, Software Architecture, Technical Leadership, OpenAPI, Back-end, Data Extraction, Product Engineering, Early-stage Startups, SaaS, Startups, Cost Modeling
Senior Data Scientist
2020 - 2022
Protium
- Built a credit card churn prediction model using XGBoost and behavioral transaction features to identify high-risk inactive borrowers and improve retention campaign targeting.
- Built a customer retention model for lending users using transaction, repayment, and engagement signals to predict repeat borrowing propensity and improve renewal conversion rates.
- Developed XGBoost and random forest models on GCP Vertex AI across 2+ million records, achieving a 76% AUC, reducing bad rates by 5%, increasing approvals by 10%, and cutting defaults by 3% in production.
- Built a reconstruction-error autoencoder on normal loan data, eliminating label dependency, reducing manual fraud reviews by 35%, lowering fraud losses by 12%, and accelerating loan approvals.
- Deployed AWS monitoring pipeline with Airflow and MLflow tracking AUC, PSI, and drift across 6 models; automated retraining triggers reduced performance incidents by 30%.
Technologies: Python, SQL, PySpark, Machine Learning, Deep Learning, Excel 2010, Metabase, Data Analytics, Data Visualization, Modeling, Regression, XGBoost, Random Forests, CatBoost, Linear Regression, Logistic Regression, Neural Networks, API Integration, REST APIs, Artificial Intelligence (AI), Data Science, Amazon Web Services (AWS), Machine Learning Operations (MLOps), Pandas, NumPy, SciPy, APIs, A/B Testing, Hypothesis Testing, Predictive Analytics, Automation, Data Pipelines, Data Collection, PostgreSQL, Data Modeling, Banking & Finance, Financial Systems, Microservices, Statistical Modeling, TensorFlow, Spark, Docker, Financial Markets, Algorithms, Time Series Forecasting, Credit Underwriting, Underwriting, Optuna, ETL, Analytics, Predictive Modeling, Microsoft PowerPoint, Communication, Performance Marketing, Text Classification, Classification, Gradient Boosting, Time Series, K-means Clustering, Dimensionality Reduction, CI/CD Pipelines, Pydantic, Pytest, Kubernetes, Data Engineering, OpenCV, Object Recognition, TensorFlow Lite, Deployment, Business Analysis, Testing, Deep Neural Networks (DNNs), NLU, ETL Pipelines, Model Validation, ARIMA, StatsModels, BigQuery, Project Management, GitHub, Microsoft Excel, Financial Modeling, API Development, ML Pipelines, Scalability, Docker Swarm, Data Scientist, Marketing Mix Modeling, Jupyter Notebook, Risk Modeling, Risk Models, ETL Tools, Real-time Data, Streaming Data, Scikit-learn, LightGBM, Hugging Face, Credit Risk, Financial Data, Fraud Detection, Financial Data Analytics, RESTFul APIs, Debugging, Probabilistic Modeling, Vertex AI, Logistics & Supply Chain, AWS IoT, API Design, Platform Design, Hyperparameter Tuning, Rule-based Programming, Asynchronous Programming, Cloud Platforms, Event-driven Architecture, Model Evaluation, Cloud, Product Development, Regression Modeling, Pricing Elasticity, Statistical Analysis, Statistical Methods, Python, Machine Learning, Product Management, Amazon Web Services (AWS), Product Development, Asyncio, Coding, Finance, Stakeholder Management, Clustering Algorithms, Prediction Markets, ETL Development, JavaScript, Django, High-performance Computing (HPC), DevOps, Bayesian Statistics, Churn Analysis, Large-scale Data Processing, Pricing Models, API Gateways, Monitoring, Amazon SageMaker, GIS, GeoPandas, Git, Model Deployment, Model Development, Datasets, Image Generation, Technical Architecture, GitLab CI/CD, BERT, SHAP, EDA, Matplotlib, Seaborn, JSON, Data Quality, Software Architecture, Back-end, Data Extraction, Early-stage Startups, Startups, Cost Modeling
Data Scientist
2019 - 2019
L&T India
- Built and fine-tuned ResNet and EfficientNet CNN models in PyTorch for construction defect detection, achieving 92% accuracy; automated inspections, reducing manual review time by 40% and improving on-site quality control.
- Trained LSTM and Transformer models on historical project data for cost and delay prediction, improving time estimation accuracy by 18% and enabling proactive risk mitigation in large-scale construction projects.
- Built a Mask R-CNN defect segmentation model in PyTorch for construction site images, achieving 90% IoU; enabled automated crack and structural anomaly detection, significantly improving on-site inspection accuracy and safety compliance.
- Performed advanced EDA using Pandas, NumPy, and PySpark on 500,000+ construction records to analyze material usage, costs, and timelines; improved forecasting accuracy by 15% and reduced budget overruns through data-driven planning.
Technologies: Python, Data Analysis, Data Visualization, Excel 2010, SQL, Deep Learning, Machine Learning, Computer Vision, YOLOv5, Optical Character Recognition (OCR), Convolutional Neural Networks (CNNs), Linear Regression, Logistic Regression, XGBoost, API Integration, REST APIs, Artificial Intelligence (AI), Data Science, Agentic Frameworks, Databricks, Machine Learning Operations (MLOps), Pandas, NumPy, SciPy, Hypothesis Testing, Data Collection, Image Classification, Data Modeling, Statistical Modeling, TensorFlow, Analytics, Predictive Modeling, Microsoft PowerPoint, Communication, Gradient Boosting, Dimensionality Reduction, CI/CD Pipelines, Pydantic, Pytest, Workflow, OpenCV, Object Recognition, TensorFlow Lite, Deployment, Testing, Deep Neural Networks (DNNs), Natural Language Understanding (NLU), NLU, ETL Pipelines, Model Validation, Project Management, GitHub, Microsoft Excel, ML Pipelines, Data Scientist, Jupyter Notebook, Scikit-learn, Probabilistic Modeling, Image Segmentation, You Only Look Once (YOLO), API Design, Platform Design, Hyperparameter Tuning, Asynchronous Programming, Product Development, Regression Modeling, Pricing Elasticity, Statistical Analysis, Statistical Methods, Python, Machine Learning, Product Management, Product Development, Coding, Stakeholder Management, JavaScript, Django, Bayesian Statistics, API Gateways, Git, Technical Architecture, GitLab CI/CD, SHAP, EDA, Matplotlib, Seaborn, JSON, Data Quality, Software Architecture, Back-end, Data Extraction
Python Developer
2015 - 2017
Freelance.com
- Developed Python back-end services using Flask/Django, creating REST APIs for recommendation engines, sentiment analysis, and anomaly detection, integrating ML models (scikit-learn, NLTK) with MySQL/Redis for scalable data processing.
- Engineered back-end systems for data-driven applications, including URL shorteners, email classification, and reporting tools, leveraging REST APIs, database optimization, and basic machine learning for intelligent decision-making.
- Built microservices in Python for log analysis, task automation, and user management, implementing background jobs with Celery, caching via Redis, and exposing APIs for real-time insights and system monitoring.
Technologies: Python, APIs, FastAPI, Django, Flask, Probabilistic Modeling, API Design, Platform Design, Asynchronous Programming, Product Development, Python, Machine Learning, Product Management, Product Development, Asyncio, Coding, JavaScript, Git, Technical Architecture, GitLab CI/CD, EDA, Matplotlib, Seaborn, JSON, Data Quality, Software Architecture, Back-end, Data Extraction, Startups
Experience
Agentic RAG Outlook Assistant (Apollo Global via Turing)
Production agentic email assistant for Apollo Global using LangGraph multi-step pipelines with thread summarization, intent classification, and conditional reply-generation nodes. Azure AI Search hybrid retrieval (dense and keyword) with metadata filtering, scoped context to sender history and deal stage. Applied Presidio-based PII masking and regex guardrails pre-retrieval and post-generation, with sliding-window conversation compression to handle long thread context within token limits.
FastAPI back-end containerized on Docker, deployed to Azure App Service with Application Insights for latency tracing and error alerting. Structured output parsing enforced reply format consistency across varied email intents. Achieved around 3s end-to-end latency, 90% reduction in manual drafting workload, and ranked top three company-wide project at Turing.
FastAPI back-end containerized on Docker, deployed to Azure App Service with Application Insights for latency tracing and error alerting. Structured output parsing enforced reply format consistency across varied email intents. Achieved around 3s end-to-end latency, 90% reduction in manual drafting workload, and ranked top three company-wide project at Turing.
Multilingual Agentic Customer Support Platform (RVIN)
Agentic RAG platform for Arabic customer support using LangGraph pipelines with query rewriting, and parent-child chunking for fragmented Arabic policy docs. Hybrid retrieval combined Pinecone dense search with BM25 sparse retrieval, re-ranked via Cohere Rerank three cross-encoder scoring. Multi-hop retrieval decomposed compound Arabic queries into sub-queries, retrieved them independently, and then fused them via context-synthesis prompts.
RAGAS continuously evaluated faithfulness, relevance, and context precision to detect retrieval drift post-updates. Redis cached top-K chunks per session, reducing redundant vector lookups on follow-up turns. Reduced hallucination by 38%, improved retrieval precision by 52%, and maintained sub-60ms latency across Egyptian, Khaleeji, and Levantine query patterns.
RAGAS continuously evaluated faithfulness, relevance, and context precision to detect retrieval drift post-updates. Redis cached top-K chunks per session, reducing redundant vector lookups on follow-up turns. Reduced hallucination by 38%, improved retrieval precision by 52%, and maintained sub-60ms latency across Egyptian, Khaleeji, and Levantine query patterns.
Multi-agent Conversational Recommendation System (Turing)
https://github.com/shadabansari794/tripgearagent2Built a multi-agent chatbot using LangGraph-orchestrated pipelines with specialized nodes for intent classification, slot filling, and query decomposition. Integrated Whisper-based speech-to-text with dialect normalization for voice input, feeding into a retrieval agent backed by Pinecone vector search (text-embedding-3-large) with MMR-based diversity ranking across 10,000+ products. Parent-child chunking preserved product attribute structure, while metadata filters scoped retrieval by category, price range, and availability in real time.
Conversation memory managed via LangChain's buffer window memory maintained a multi-turn context for follow-up queries and preference tracking. A recommendation agent fused retrieved candidates with the user session history using a re-ranking prompt to surface personalized results. Hybrid dense-sparse retrieval improved recall on ambiguous or short queries common in voice input. Improved product discovery efficiency and increased user engagement by 35% through contextually grounded, conversational shopping experiences with sub-2s response latency.
Conversation memory managed via LangChain's buffer window memory maintained a multi-turn context for follow-up queries and preference tracking. A recommendation agent fused retrieved candidates with the user session history using a re-ranking prompt to surface personalized results. Hybrid dense-sparse retrieval improved recall on ambiguous or short queries common in voice input. Improved product discovery efficiency and increased user engagement by 35% through contextually grounded, conversational shopping experiences with sub-2s response latency.
Multi-agent AI Customer Support Platform using Amazon Bedrock
Designed and built a production-ready multi-agent AI platform to automate customer support queries using Amazon Bedrock, LangGraph, and LangChain. The system uses an orchestrator agent that breaks down user queries and routes them to specialized agents, including retrieval agents (RAGs with Bedrock Knowledge Bases), reasoning agents powered by Bedrock foundation models, and tool agents that call external APIs (order status, ticket creation, account lookup).
The platform integrates vector retrieval from documents stored in Amazon S3, contextual conversation memory, and scalable serverless infrastructure using AWS Lambda, API Gateway, and DynamoDB. The solution enables accurate knowledge-grounded responses and automated workflows, significantly improving response time and reducing manual support workload.
The platform integrates vector retrieval from documents stored in Amazon S3, contextual conversation memory, and scalable serverless infrastructure using AWS Lambda, API Gateway, and DynamoDB. The solution enables accurate knowledge-grounded responses and automated workflows, significantly improving response time and reducing manual support workload.
Credit Card Marketing Mix Model (MMM)
Built a credit card Marketing Mix Model using Bayesian regression (PyMC3) with hierarchical priors and LightGBM ensemble to measure the incremental impact of marketing channels on customer acquisitions. Applied Adstock transformations with geometric decay for carryover effects and Hill function saturation curves to model diminishing returns per channel. Decomposed revenue contributions using Shapley-based attribution to isolate baseline versus paid media lift across digital and offline touchpoints.
Integrated media spend, promotional calendars, and seasonal signals (Fourier terms for cyclical patterns) as exogenous regressors to improve forecast accuracy. Estimated channel-level ROI with posterior credible intervals to quantify uncertainty in budget recommendations. Built a constrained optimization layer using SciPy to simulate budget reallocation scenarios under spend caps, surfacing optimal channel mix for maximum acquisition yield. Model validation via MAPE and out-of-sample backtesting ensured reliability across campaign cycles.
Integrated media spend, promotional calendars, and seasonal signals (Fourier terms for cyclical patterns) as exogenous regressors to improve forecast accuracy. Estimated channel-level ROI with posterior credible intervals to quantify uncertainty in budget recommendations. Built a constrained optimization layer using SciPy to simulate budget reallocation scenarios under spend caps, surfacing optimal channel mix for maximum acquisition yield. Model validation via MAPE and out-of-sample backtesting ensured reliability across campaign cycles.
Credit Risk & Loan Default Prediction System
Developed a production-grade credit risk system using stacked XGBoost and CatBoost ensembles with Platt scaling for calibrated default probabilities. Engineered 3,000+ behavioral, transactional, geolocation, and SMS-derived features in PySpark—including repayment velocity, spending entropy, and NLP-extracted payment signals from unstructured SMS data. Applied WOE binning, IV-based feature selection, and target encoding for high-cardinality variables. Validated across temporal cohorts using AUC, KS statistic, and Gini coefficient, achieving around 75% AUC via out-of-time validation.
Deployed as a real-time FastAPI scoring endpoint with sub-100ms latency, containerized on Docker with CI/CD versioned rollouts. Integrated SHAP explainability for per-prediction attribution to meet regulatory interpretability requirements. Monitored production stability via PSI for covariate drift, CSI for score distribution shifts, and AUC decay alerting. Score-cutoff optimization aligned to business risk appetite improved loan approval rates while reducing bad rates across borrower segments.
Deployed as a real-time FastAPI scoring endpoint with sub-100ms latency, containerized on Docker with CI/CD versioned rollouts. Integrated SHAP explainability for per-prediction attribution to meet regulatory interpretability requirements. Monitored production stability via PSI for covariate drift, CSI for score distribution shifts, and AUC decay alerting. Score-cutoff optimization aligned to business risk appetite improved loan approval rates while reducing bad rates across borrower segments.
TelegramRAGBot
https://github.com/shadabansari794/TelegramRAGBotA RAG chatbot built for Telegram that allows users to query and interact with past messages or posts from specific channels. It integrates OpenAI GPT models with vector search to provide context-aware, factual responses grounded in historical Telegram data.
I designed and developed the complete pipeline, from data ingestion and cleaning to vector embedding and retrieval using FAISS. The system stores channel messages, converts them into semantic embeddings, and retrieves the most relevant context at query time for answer generation.
I implemented prompt engineering techniques to reduce hallucinations and enhance response relevance while optimizing for latency and scalability. The bot was deployed using FastAPI with asynchronous message handling, enabling near real-time interactions.
This project demonstrates production-ready RAG design, modular agent architecture, and efficient integration between LLMs, vector databases, and messaging APIs.
I designed and developed the complete pipeline, from data ingestion and cleaning to vector embedding and retrieval using FAISS. The system stores channel messages, converts them into semantic embeddings, and retrieves the most relevant context at query time for answer generation.
I implemented prompt engineering techniques to reduce hallucinations and enhance response relevance while optimizing for latency and scalability. The bot was deployed using FastAPI with asynchronous message handling, enabling near real-time interactions.
This project demonstrates production-ready RAG design, modular agent architecture, and efficient integration between LLMs, vector databases, and messaging APIs.
Credit Card Customer Churn Prediction Model
Developed a machine learning-based credit card churn prediction system using XGBoost and behavioral analytics on large-scale customer transaction data. Engineered features from spending behavior, repayment trends, credit utilization, transaction frequency, app engagement, delinquency patterns, and customer service interactions to identify customers at high risk of attrition. Built scalable data pipelines using PySpark and SQL for processing millions of records and deployed the model for proactive retention targeting, personalized offers, and cross-sell optimization. The solution helped improve customer retention strategies, reduce inactive user rates, and optimize marketing campaign efficiency through risk-based segmentation and predictive analytics.
Automated Billing & Budget Workflow Platform
Designed and developed a centralized billing and budget management platform that replaced spreadsheet-driven processes with automated workflows and real-time financial tracking. Built a React-based spreadsheet-style interface, FastAPI back-end services, and PostgreSQL database to manage budgets, invoices, approvals, audit history, and reporting.
Leveraged n8n as the workflow orchestration engine to automate budget approvals, invoice generation, payment tracking, exception handling, and stakeholder notifications. Integrated QuickBooks for accounting operations, Airtable for operational planning, Google Workspace, and Slack/email notifications to create a seamless end-to-end workflow.
Implemented role-based access controls, configurable approval chains, validation rules, and audit logging to ensure compliance and data accuracy. Developed automated reconciliation workflows that synchronized financial records across systems, reduced manual effort by over 70%, and improved visibility into budget utilization and billing status. Deployed the solution on AWS with monitoring, logging, and CI/CD pipelines, enabling scalable and reliable operations across multiple business teams.
Leveraged n8n as the workflow orchestration engine to automate budget approvals, invoice generation, payment tracking, exception handling, and stakeholder notifications. Integrated QuickBooks for accounting operations, Airtable for operational planning, Google Workspace, and Slack/email notifications to create a seamless end-to-end workflow.
Implemented role-based access controls, configurable approval chains, validation rules, and audit logging to ensure compliance and data accuracy. Developed automated reconciliation workflows that synchronized financial records across systems, reduced manual effort by over 70%, and improved visibility into budget utilization and billing status. Deployed the solution on AWS with monitoring, logging, and CI/CD pipelines, enabling scalable and reliable operations across multiple business teams.
Multimodal Conversational AI System
I designed and developed a production-ready multimodal conversational AI platform delivering personalized, context-aware text and voice interactions. I architected an LLM orchestration and RAG pipeline using Claude APIs, Meta Llama, prompt engineering, and Pinecone vector memory to retain long-term semantic context and user preferences.
My work included integrating Whisper for speech-to-text, TTS for voice responses, and generative image models for multimodal experiences. I built scalable data pipelines for chat datasets, embeddings, user profiles, and model evaluation, exposing capabilities via FastAPI endpoints with Rasa NLU for intent classification. I also benchmarked third-party LLM APIs against fine-tuned open-source models on latency, cost, and reasoning quality.
Finally, I applied neural networks, tree-based models, and reinforcement learning to improve personalization and response selection. I also established LLMOps workflows covering evaluation, monitoring, safety guardrails, and production deployment.
My work included integrating Whisper for speech-to-text, TTS for voice responses, and generative image models for multimodal experiences. I built scalable data pipelines for chat datasets, embeddings, user profiles, and model evaluation, exposing capabilities via FastAPI endpoints with Rasa NLU for intent classification. I also benchmarked third-party LLM APIs against fine-tuned open-source models on latency, cost, and reasoning quality.
Finally, I applied neural networks, tree-based models, and reinforcement learning to improve personalization and response selection. I also established LLMOps workflows covering evaluation, monitoring, safety guardrails, and production deployment.
Title Production Spend Forecasting & Rigorous Time-series Backtesting
Designed and developed a production-grade forecasting framework to predict future member-level spend and utilization from large-scale longitudinal claims data. Engineered temporal, recency, frequency, utilization, diagnosis, service-category, and historical spend features using PySpark and SQL, and trained LightGBM/XGBoost models for forward-looking cost prediction. Built a rigorous time-aware backtesting framework using rolling and expanding validation windows to ensure every model was evaluated strictly on future, unseen periods and to prevent temporal leakage. Compared performance across forecast horizons, member cohorts, and high-cost segments using MAE, RMSE, calibration, rank ordering, and stability metrics rather than relying only on training-set fit. Used SHAP for model interpretation and MLflow for experiment tracking, reproducibility, model comparison, and production monitoring. The framework supported model selection based on consistent out-of-sample performance and real-world stability rather than a single validation split.
Education
2017 - 2019
Master's Degree in Mechanical Engineering
IIT Kharagpur - Kharagpur, India
2011 - 2015
Bachelor's Degree in Mechanical Engineering
Haldia Institute of Technology - India
Certifications
FEBRUARY 2026 - PRESENT
Academy Accreditation - Generative AI Fundamentals
Databricks
JANUARY 2026 - PRESENT
AWS Certified Cloud Practitioner
AWS
JULY 2025 - PRESENT
Knowledge Graphs for RAG
DeepLearning.AI
JULY 2025 - PRESENT
MCP: Build Rich-context AI Apps with Anthropic
DeepLearning.AI
MARCH 2025 - PRESENT
Serverless Agentic Workflows with Amazon Bedrock
DeepLearning.AI
FEBRUARY 2025 - PRESENT
LangChain for LLM Application Development
DeepLearning.AI
FEBRUARY 2025 - PRESENT
Introducing Multimodal Llama 3.2
DeepLearning.AI
NOVEMBER 2019 - PRESENT
Data Structure and Algorithms
Udemy
AUGUST 2018 - PRESENT
Time Series Analysis
Coursera
AUGUST 2018 - PRESENT
Machine Learning
Coursera
Skills
Libraries/APIs
PySpark, XGBoost, CatBoost, PyTorch, Scikit-learn, REST APIs, Pandas, NumPy, SciPy, OpenAI API, TensorFlow, Pydantic, Auth, API Development, Beautiful Soup, WhatsApp API, React, Asyncio, Hugging Face Transformers, Imbalanced-learn, Node.js, Rasa NLU, OpenAPI, OpenCV, Claude API, GraphQL API, vLLM, Matplotlib
Tools
Apache Airflow, Excel 2010, ChatGPT, Microsoft PowerPoint, Pytest, AWS Glue, Microsoft Copilot, Amazon Textract, ARIMA, StatsModels, BigQuery, GitHub, Microsoft Excel, Claude Code, Claude, AI SDK, Amazon ElastiCache, Claude Agent SDK, n8n, Azure Kubernetes Service (AKS), Azure Machine Learning, Azure ML Studio, Codex, Git, Amazon EKS, Whisper, Rasa.ai, Docker Swarm, Azure OpenAI Service, Terraform, Observability Tools, You Only Look Once (YOLO), Visual Language Models (VLMs), Amazon SageMaker, Open Neural Network Exchange (ONNX), GraphRAG, GIS, RingCentral, Grafana, GitLab CI/CD, Seaborn
Languages
Python, SQL, Python 3, SAS, Snowflake, JavaScript, TypeScript, Go, C#, Java
Frameworks
Flask, LangGraph, LightGBM, Agentic Frameworks, Spark, Optuna, TensorFlow Lite, AutoGen, Django, LlamaIndex, Streamlit, Next.js, .NET, ASP.NET
Paradigms
Model Context Protocol (MCP), Automation, Microservices, ETL, Testing, Rule-based Programming, Asynchronous Programming, Event-driven Design (EDD), Synthetic Data Generation, High-performance Computing (HPC), HIPAA Compliance, Event-driven Architecture, Business Intelligence (BI), DevOps
Platforms
AWS Lambda, Amazon Web Services (AWS), Azure, Google Cloud Platform (GCP), Docker, Kubernetes, Amazon EC2, Jupyter Notebook, Azure AI Search, Microsoft Copilot Studio, Azure AI Studio, Vertex AI, AWS IoT, LangSmith, Twilio, Vercel, Ollama, Harness, Langfuse, Palantir Foundry, Azure Functions, Replit, Cloud Run, Databricks, Microsoft Power Platform, CrewAI, Observable Framework, Kubeflow, Cortex
Storage
Amazon S3 (AWS S3), Data Pipelines, PostgreSQL, Redis, Neo4j, Graph Databases, MongoDB, Datadog, JSON, Data Lakes, Azure Cosmos DB
Industry Expertise
Banking & Finance, Project Management, Healthcare, Marketing, Bioinformatics
Other
Machine Learning, Deep Learning, Time Series Analysis, Statistics, Retrieval-augmented Generation (RAG), Agentic AI, Chatbots, Natural Language Processing (NLP), Generative Artificial Intelligence (GenAI), Model Monitoring, Feature Engineering, Metabase, Data Analytics, Data Visualization, Modeling, Regression, Random Forests, Linear Regression, Logistic Regression, Neural Networks, Data Analysis, Computer Vision, YOLOv5, Optical Character Recognition (OCR), Convolutional Neural Networks (CNNs), FastAPI, LangChain, Prompt Engineering, Large Language Models (LLMs), OpenAI GPT-4 API, Conversational AI, Forecasting, Meta Llama, OpenAI GPT-3 API, Knowledge Graphs, API Integration, OpenAI, Artificial Intelligence (AI), Data Science, Machine Learning Operations (MLOps), Pinecone, APIs, A/B Testing, Data Analytics (Marketing), Marketing Analytics, Hypothesis Testing, Predictive Analytics, Marketing Mix Modeling, Data Collection, Web Scraping, Predictive Maintenance, Architecture, AI Modeling, AI Model Training, Vector Databases, eCommerce, Anthropic, Sentiment Analysis, AI Automation, ChatGPT API, Data Modeling, AI Tools, AI Chatbots, FAISS, Document Processing, Large Language Model Operations (LLMOps), Financial Systems, Document Parsing, Statistical Modeling, MLflow, Hugging Face, Financial Markets, Algorithms, Time Series Forecasting, Credit Underwriting, Underwriting, Analytics, Predictive Modeling, Marketing Attribution, Communication, Performance Marketing, Text Classification, Minimum Viable Product (MVP), Classification, Decision Trees, Gradient Boosting, Time Series, K-means Clustering, Dimensionality Reduction, CI/CD Pipelines, Data Engineering, Data Handling, Workflow, Deployment, Llama 3, Object Detection, Business Analysis, AI Consulting, AI Design, OAuth, Generative Pre-trained Transformers (GPT), Tesseract, Chatbot Conversation Design, Open-source LLMs, Speech-to-Text (STT), Text-to-Speech (TTS), RAG Pipelines, AI Assistants, Agentic RAG Systems, Deep Neural Networks (DNNs), Natural Language Understanding (NLU), NLU, RAG Architecture, ETL Pipelines, Model Validation, AI Copilots, Financial Modeling, AI-generated Video, Workflows, Data Privacy, ML Pipelines, Scalability, Semantic Search, Data Governance, Data Scientist, AI Architecture, Risk Modeling, Risk Models, Website Data Scraping, ETL Tools, RAG Systems, Custom Models, Credit Risk, Financial Data, Financial Data Analytics, RESTFul APIs, Foundry, Low Code, Financial Analysis, AI Pipeline, AI Agents, Debugging, API Design, Platform Design, Scraping, Probabilistic Modeling, Logistics & Supply Chain, Consulting, Scalable Vector Databases, Hyperparameter Tuning, Benchmarking, Cloud Platforms, Multi-agent Systems, Model Evaluation, System Architecture, Supabase, ElevenLabs Solutions, Cloud, Product Development, Regression Modeling, Statistical Analysis, Statistical Methods, AI Agents, Agentic AI, Artificial Intelligence (AI), Generative Artificial Intelligence (GenAI), Large Language Models (LLMs), Python, Machine Learning, Product Management, Amazon Web Services (AWS), Azure, Product Development, System Design, OpenAI SDK, Platform Engineering, Technical Writing, AI Marketing, Coding, Gemini, LoRa, QLoRA, BERT, SHAP, EDA, Leadership, Team Leadership, SDKs, Finance, Vector Search, Unity Catalog, Agentic AI Systems, Governance, Regression Testing, Cloud Architecture, Data Architecture, Enterprise Architecture, Solution Architecture, Stakeholder Management, Decision Modeling, Clustering Algorithms, Recommendation Systems, Prediction Markets, Data Warehousing, Embedding Models, ETL Development, Azure AI Services, AI Agent Orchestration, Light LLMs, Pgvector, UiPath, Bayesian Statistics, Multiagent Generative Systems (MAGs), LLM Integration, Semantic Code, Churn Analysis, Large-scale Data Processing, Cloud Governance, API Gateways, Agentic Coding, Audio Processing, Automatic Speech Recognition (ASR), Distributed Systems, Monitoring, Cursor AI, Industrials, Azure Function App, Amazon Bedrock AgentCore, Full-stack, Application State Management, GeoPandas, Data Security, Quantization, AI Hallucinations Management, Agentic Workflow Design, HIPAA, Qwen, AI platform engineer, Machine Learning Infrastructure Engineer, Demand Forecasting, Model Deployment, Model Development, Datasets, LLM Fine-tuning, Documentation, LLM Reasoning, Image Generation, Multimodal GenAI, Small Language Models (SLMs), Technical Architecture, Data Quality, Delta Lake, Apollo, Clay, Text to Image, Text to Image AI, Software Architecture, Technical Leadership, Back-end, Data Extraction, Google Document AI, Handwriting Recognition, Applied AI, Early-stage Startups, SaaS, Startups, Combinatorial Optimization, Cost Modeling, LLM Agents, OpenAI Agents SDK, AI Model Integration, Image Classification, Multimodal Models, Reinforcement Learning, Scientific Data Analysis, Object Recognition, Workflow Automation, Reliability, Spatial Analysis, AI Voice Agents, Google Calendar, Real-time Data, Fraud Detection, Reporting, Observability, CRM APIs, Workflow Automation & System Integration, WhatsApp, Image Segmentation, Webhooks, Qdrant, ColBERT, Pricing Elasticity, Pricing Models, Artificial Intelligence as a Service (AIaaS), iGaming, Clinical Research, ONNX Runtime, Graphics Processing Unit (GPU), Fine-tuning, Azure Databricks, Telemetry, Product Engineering, Content, Azure AI Document Intelligence, Streaming Data, Semantic Kernel (SK), NVIDIA TensorRT, Airtable, Transformers
Collaboration That Works
How to Work with Toptal
Toptal matches you directly with global industry experts from our network in hours—not weeks or months.
1
Share your needs
Discuss your requirements and refine your scope in a call with a Toptal domain expert.
2
Choose your talent
Get a short list of expertly matched talent within 24 hours to review, interview, and choose from.
3
Start your risk-free talent trial
Work with your chosen talent on a trial basis for up to two weeks. Pay only if you decide to hire them.
Top talent is in high demand.
Start hiring