Divyashakti Kumar
Verified Expert in Engineering
AI Developer
Bengaluru, Karnataka, India
Toptal member since June 26, 2025
Divyashakti is a senior AI engineer with over 8 years of experience leading the shift from predictive analytics to Agentic AI and autonomous systems. He specializes in designing multi-agent workflows, generative AI applications, and advanced NLP/vision pipelines. With a proven track record in leveraging scalable LLMOps, he deploys reasoning-driven, enterprise-grade AI solutions that automate complex operations, optimize decision-making, and deliver immediate business value.
Portfolio
Experience
- Data Science - 6 years
- PyTorch - 6 years
- Python - 6 years
- Computer Vision - 5 years
- Machine Learning Operations (MLOps) - 4 years
- Natural Language Processing (NLP) - 4 years
- Artificial Intelligence (AI) - 3 years
- LangGraph - 2 years
Preferred Environment
Visual Studio Code (VS Code), Python, PyTorch, TensorFlow, Hugging Face Transformers, Computer Vision, Natural Language Processing (NLP), Artificial Intelligence (AI), Machine Learning Operations (MLOps), Amazon Bedrock AgentCore
The most amazing...
...thing I've developed is a personal assistant that handles all my research and completes my to-dos, so I can focus on what truly matters.
Work Experience
AI Engineer
Maneva, Inc.
- Owned the end-to-end delivery of a computer vision project, independently handling client requirements, data preparation, model development, deployment, and production integration.
- Designed an automated, agentic model feedback loop that monitored model performance and triggered retraining workflows when performance degraded, enabling continuous model improvement with minimal manual intervention.
- Managed the complete ML/CV project lifecycle for the client, from requirement analysis and data pipelines to model optimization, deployment, monitoring, and continuous improvements, ensuring the solution met business and technical requirements.
- Developed and optimized computer vision models to achieve strong accuracy, robustness, and reliability across real-world use cases and varying input conditions.
Lead Generation AI Engineer
Micoworks
- Created an agentic pipeline for automated ad-creatives generation.
- Developed a pipeline containing agents to generate product-based prompts, facilitating the creation of background and foreground images.
- Built complex RAG pipelines and agentic frameworks, achieving one second latency for search query.
Senior Data Scientist
Augmentolabs Digital
- Trained YOLOv8 model for precisely detecting bedroom and kitchen furniture and objects, achieving an impressive 0.78 mAP score.
- Expanded the model pipeline to facilitate item search within catalog data, seamlessly integrating it into the existing framework and deploying the entire solution effectively.
- Created a pipeline for generating and editing room interior images, incorporating JSON file-based prompt conditioning for image generation.
Data Scientist
Acviss Pvt
- Designed and modified QR codes to make them unreadable by other QR scanners for counterfeit tracing.
- Developed a platform to analyze user sentiment related to brands using feeds shared on social media.
- Designed and developed a Docker-based inference engine for scalable model development and deployment.
Data Scientist
DeepQuanty AI Labs
- Automated the process of cheque verification with 85% efficiency and lowered manual verification to 15%.
- Implemented object detection architectures like YOLOv3, RetinaNet, EfficientDet, and MORAN to extract data from financial documents and entity mapping.
- Automated a data extraction process from financial documents, yielding an 80% reduction of work per document.
Experience
Super Agent
Vision-based Inspection System
TECHNOLOGIES USED
• Back end: Python, FastAPI
• Computer Vision: OpenCV, TensorFlow/Keras
• Image Processing: NumPy, PIL
• Database: PostgreSQL, SQLAlchemy
• Deployment: Docker, REST APIs
KEY FEATURES
• Real-time defect detection (scratches, dents, and burnt marks )
• Automated pass/fail classification
• Multi-camera integration
• WebSocket live monitoring
• Batch processing capabilities
ACHIEVEMENTS
• 99.5% accuracy in defect detection
• 2 seconds inspection time per coil
• 1,200+ coils/hour throughput capacity
• <2% false positive rate
• 50,000+ images training dataset
• 95% test coverage, comprehensive testing
BUSINESS IMPACT
Eliminated manual inspection errors, increased production efficiency by 400%, and reduced quality control costs by 60% while maintaining consistent inspection standards.
AI Voice Interview Agent
CORE TECHNOLOGIES
Speech Processing: Speech-to-text (Google and Azure), text-to-speech synthesis
AI Models: OpenAI GPT, Google Gemini, DeepSeek AI
Real-time Communication: LiveKit WebRTC
Back end: FastAPI, WebSocket connections
Audio Processing: PyAudio, librosa
KEY FEATURES
• Natural conversation flow with context awareness
• Multi-language support for global candidates
• Real-time audio streaming with <200ms latency
• Dynamic question generation based on responses
• Automated scoring with detailed feedback
• Interview recording and transcription
TECHNICAL IMPLEMENTATION
• Bi-directional audio streaming via LiveKit
• LLM-powered question adaptation and follow-ups
• RESTful APIs for interview management
• Real-time WebSocket for live monitoring
ACHIEVEMENTS
• 95% speech recognition accuracy
• < 2-second response time end-to-end
• 500+ concurrent interview sessions
• Multi-modal evaluation (speech, content, sentiment)
• Reduced hiring time by 70% with consistent evaluation standards
Education
Bachelor's Degree in Electrical Engineering
BIT Sindri - Dhanbad, India
Skills
Libraries/APIs
PyTorch, Hugging Face Transformers, NumPy, OpenAI API, OpenCV, API Development, TensorFlow, WebRTC, Keras
Tools
You Only Look Once (YOLO), Open Neural Network Exchange (ONNX), AI Prompts, Claude, GraphRAG
Languages
Python, SQL, SPARQL
Frameworks
LlamaIndex, LangGraph, Agentic Frameworks, AutoGen, Next.js, Optuna
Paradigms
DevOps, ETL, Model Context Protocol (MCP)
Platforms
Docker, Jupyter Notebook, Firebase, Azure AI Studio, Harness, Linux, AWS IoT, Amazon Web Services (AWS), Kubernetes, CrewAI, Apache Kafka, Azure, Google Cloud Platform (GCP), Visual Studio Code (VS Code), Vertex AI, LiveKit
Storage
Amazon S3 (AWS S3), Cloud Firestore, Data Pipelines, Redis
Other
Computer Vision, Natural Language Processing (NLP), Artificial Intelligence (AI), Data Science, FastAPI, Computer Vision Algorithms, APIs, Image Processing, Text-to-Speech (TTS), Large Language Models (LLMs), Retrieval-augmented Generation (RAG), Machine Learning, AI Agents, Generative Pre-trained Transformers (GPT), Transformers, Hugging Face, Data Analytics, Data Analysis, Predictive Modeling, Video Processing, Image Segmentation, Image Analysis, OpenAI, Architecture, Prompt Engineering, Software Architecture, Audio, Game Analytics, Minimum Viable Product (MVP), Vector Databases, Medical Imaging, Convolutional Neural Networks (CNNs), API Integration, CI/CD Pipelines, Fine-tuning, ML Pipelines, Residual Neural Networks (ResNets), Image Classification, Fintech, AI Design, Team Leadership, Technical Hiring, Interviewing, Code Review, Task Analysis, Statistical Analysis, Small Language Models (SLMs), Containerization, MLflow, Vision Transformer (ViT), Optical Character Recognition (OCR), Amazon Bedrock AgentCore, AI Architecture, Context Engineering, RAG Systems, AI Compliance, AI Security, Vibe Coding, ChatGPT Prompts, AI Workflow, LLM Agents, AI Voice Agents, ElevenLabs Solutions, Scalable Platforms, AI Engineering, AI Tools, Software Development Lifecycle (SDLC), AI-assisted Development, Technical Program Management, Machine Learning Operations (MLOps), LangChain, Generative Artificial Intelligence (GenAI), Agentic AI, 3D Image Processing, Knowledge Graphs, Ontologies, Microsoft Azure, Power BI, Google ADK, Object Detection, Deep Learning, Speech-to-Text (STT), Classification
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