Sehar Nisar
Verified Expert in Engineering
Machine Learning Engineer and Developer
Lahore, Pakistan
Toptal member since August 4, 2026
Sehar is a senior machine learning engineer with six years of experience delivering enterprise-grade AI systems for clients, including Azunex, Pfizer, and SCFS. She specializes in large language models, RAG, and agentic AI across healthcare, pharmaceutical, and surveillance industries. At Azunex, Sehar built a multi-stage LangGraph research agent, and at Pfizer, she designed agentic reasoning systems for multi-step orchestration and compliance workflows.
Portfolio
Experience
- Python - 6 years
- Text Classification - 4 years
- LangChain - 3 years
- Embeddings - 3 years
- LangGraph - 3 years
- Retrieval-augmented Generation (RAG) - 3 years
- Vector Databases - 3 years
- Large Language Models (LLMs) - 3 years
Preferred Environment
Amazon Web Services (AWS), Google Cloud Platform (GCP), Agentic AI, CI/CD Pipelines, Embeddings, Feature Engineering, LangChain, LangGraph, Large Language Models (LLMs), Retrieval-augmented Generation (RAG)
The most amazing...
...work I've done is redesigning a multi-agent voice bot for taxi booking, cutting cost per call from $1.50 to $0.04 (a 97% drop) while eliminating context drift.
Work Experience
Senior Machine Learning Engineer
Azunex
- Deployed a RAG pipeline behind WhatsApp and Messenger webhooks using semantic search to ground responses in real time, cutting data ingestion lag by 40%.
- Built a GCP/Vertex AI pipeline that gates model promotion from training to serving behind automated evaluation checks, letting validated models reach production without manual handoff.
- Constructed a two-layer pre-deployment evaluation system, achieving 98% test coverage for a voice AI assistant using static judge prompt analysis and a WebRTC caller agent simulating full conversations.
- Implemented a fallback routing layer that detects low-confidence agent responses mid-conversation and hands off to a simpler rules-based flow, preventing dead-end conversations during outages or edge cases.
- Built a semantic caching layer in front of the RAG pipeline that reuses embeddings for near-duplicate queries, cutting redundant vector search calls and shaving response latency.
- Built a multi-stage LangGraph research agent with Router, Planner, Researcher, Writer, and Quality Check stages, each validating its output before handoff.
Machine Learning Engineer
Pfizer
- Built a medical literature summarization tool that ingests new PubMed/clinical research and surfaces only papers relevant to active drug programs, filtered by a relevance classifier trained on past research team feedback.
- Designed agentic reasoning systems with LangGraph, implementing multi-step orchestration, tool-use patterns, and human-in-the-loop controls for compliance tasks.
- Built an internal knowledge-base chatbot for R&D teams that answers questions about prior experiments and lab results by retrieving from structured lab notebooks and unstructured PDFs.
- Orchestrated AI infrastructure via Docker/Kubernetes with observability for model drift and cost-latency trade-offs, reducing cloud compute spend by 22%.
- Built a regulatory submission QA system that checks draft FDA/EMA submission documents against formatting and content requirements automatically, catching compliance gaps before human review.
Machine Learning Engineer
SCFS
- Conducted applied research on deep learning models for video surveillance and anomaly detection, improving prediction accuracy by 14%.
- Built a real-time multi-camera object detection and tracking pipeline that flagged unauthorized access events across a client’s camera network, cutting manual monitoring hours.
- Managed multimodal datasets and 3D neural network architectures to optimize high-volume pipelines by 20%.
- Deployed optimized ML models on Raspberry Pi edge devices, cutting inference processing times by 22%.
- Built reliable CRUD APIs for a financial application, reducing runtime errors by 15% through automated testing.
- Quantized and optimized detection models for edge deployment, balancing accuracy against the compute limits of low-power surveillance hardware.
- Designed the data ingestion pipeline that pulled, labeled, and versioned video footage from multiple camera feeds for model training, replacing a manual labeling process.
Experience
Intelligent Multi-agent RAG Ecosystem
Hotel Operations and Guest Service Agent — Hospitality
• Built a property-knowledge and RAG pipeline using hotel policies, FAQs, room details, services, reservations, maintenance records, and guest reviews.
• Implemented specialized agents for guest service, maintenance, housekeeping, reservations, and concierge recommendations, coordinated through a central workflow agent.
• Integrated structured APIs/tools for reservation lookup, room status, housekeeping, maintenance tickets, and service availability, with confirmation required for sensitive write operations.
• Added state management, idempotency, authorization, rate limiting, prompt-injection protection, audit logging, structured outputs, and fallback handling for unavailable systems.
-Deployed with Next.js, FastAPI, PostgreSQL, Redis, vector search, event queues, Docker/Kubernetes, API gateway, and OpenTelemetry/Prometheus.
Contract Intelligence and Negotiation Copilot
• Developed document intelligence and RAG pipelines with OCR, clause segmentation, embeddings, hybrid search, reranking, and structured extraction.
• Implemented policy-aware risk analysis and negotiation agents for clause recommendations and lawyer-reviewable redlines.
• Added security guardrails, audit logging, RBAC, human approval workflows, and deployed using FastAPI, PostgreSQL, OpenSearch/pgvector, Docker, and Kubernetes.
Education
Bachelor's Degree in Computer Science
Punjab University - Lahore, Pakistan
Skills
Libraries/APIs
PyTorch, Claude API, Google API, REST APIs, Scikit-learn, TensorFlow, Pandas, NumPy, Hugging Face Transformers, React, Google Places API, WebRTC, Keras, SciPy
Tools
Claude, Terraform, Apache Airflow
Languages
Python, SQL, TypeScript
Frameworks
LangGraph, DSPy, Flask, Django, Next.js
Platforms
LangSmith, Langfuse, Amazon Web Services (AWS), Google Cloud Platform (GCP), Vertex AI, Docker, Kubernetes, Raspberry Pi, Azure
Storage
MySQL, PostgreSQL, Redis, MongoDB
Paradigms
Microservices, Model Context Protocol (MCP), ETL, Asynchronous Programming
Other
Retrieval-augmented Generation (RAG), Model Evaluation, Machine Learning, Natural Language Processing (NLP), Computer Vision, Text-to-text Transfer Transformer (T5), Large Language Models (LLMs), Agentic AI, AI Agents, Prompt Engineering, Artificial Intelligence (AI), Agentic RAG Systems, Full-stack Development, Data Analysis, Education Technology (Edtech), Personally Identifiable Information (PII), Software Architecture, Recommendation Systems, Prompt Optimization, A/B Testing, AI Voice Agents, Agentic AI Systems, APIs, Evaluation, FastAPI, Feature Engineering, Convolutional Neural Networks (CNNs), OpenAI GPT-4 API, Fine-tuning, LangChain, Vector Databases, Embeddings, CI/CD Pipelines, System Design, RAG Systems, RAG Architecture, Workflow Automation, PDF, Email Automation, Applied AI, Knowledge Graphs, Customer Relationship Management (CRM), Data Annotation, Text-to-Speech (TTS), DeepEval, Braintrust, Data Science, Sequence Models, Transformers, Amazon Bedrock AgentCore, ElevenLabs Solutions, AWS WAF, Agentic Workflow Design, Data-informed Recommendations, IT Consulting, Solution Architecture, Consulting, Performance Optimization, Software Engineering, Semantic Search, Deep Learning, GitHub Actions, ELT, Recurrent Neural Networks (RNNs), GAN, Llama 3, Mistral AI, Pgvector, LoRa, QLoRA, PEFT, NER, Text Classification, Deployment, Reranking, Hybrid Search, DocumentDB, Optical Character Recognition (OCR), Embedding Models, Data Processing, Classification, Healthcare IT, Explainable Artificial Intelligence (XAI), Machine Learning Operations (MLOps)
How to Work with Toptal
Toptal matches you directly with global industry experts from our network in hours—not weeks or months.
Share your needs
Choose your talent
Start your risk-free talent trial
Top talent is in high demand.
Start hiring