Open to opportunities

Prafulla Purohit

@prafullapurohit

Message

AI/ML and full-stack engineer focused on building GenAI, RAG, and production-ready applications.

India

Message

What I'm looking for

I’m looking to contribute to a Content Engineering role—building GenAI and AI-powered pipelines (RAG, multi-agent orchestration, and structured extraction) with strong analytical thinking and clear communication, shipping reliable products end-to-end.

I’m a final-year B.Tech Information Technology student who’s built end-to-end AI applications—from full-stack platforms to production-grade RAG pipelines. I enjoy turning messy, real-world inputs into structured outputs, using systems like ATS resume parsing, grounded search with citations, and multi-document synthesis with faithfulness checks.

In my internships, I’ve shipped ML and GenAI features, including real-time criminal face detection using CNN/OpenCV with embeddings and similarity matching, and SaaS training where I deployed an invoice/bill RAG pipeline with OCR, semantic chunking, embedding generation, and cached processing. I also build multi-agent workflows with LangChain/LangGraph and agentic tool-use, and I care deeply about reliability, orchestration quality, and clear technical communication.

Experience

Work history, roles, and key accomplishments

Current

AI/ML Engineer Intern

Current

Satva Solutions

Jan 2026 - Present (5 months)

Completed full-stack SaaS development training across React, Next.js, Node.js, Express, PostgreSQL, and REST APIs, shipping features on 3+ live production projects with CI/CD and agile sprints. Built and deployed a production RAG pipeline for invoice and bill processing using OCR (Tesseract/AWS Textract) and LLM-powered structured extraction for automated Q&A and reconciliation.

React Next.js Express PostgreSQL OCR (Tesseract AWS Textract)LangGraph Docker NodeJS

AI/ML Intern

TSS LTD

Jan 2025 - Jun 2025 (5 months)

Built a criminal face detection and categorization system using CNN, OpenCV, MediaPipe, and face-embedding similarity to enable real-time facial recognition and identity tagging. Developed end-to-end ML pipelines with data augmentation, fine-tuning, hyperparameter tuning, and model evaluation in TensorFlow/Keras and scikit-learn.