Samit Pawar — AI Engineer · Senior Python Developer · 8+ yrs

I'm an AI Engineer and Senior Python Developer with 8+ years building production systems — the last 2 spent redesigning how I ship code around LLM agents, RAG, and the Model Context Protocol. By day, I lead the AI-native engineering platform behind Trulla, a 340B healthcare-compliance SaaS serving 30+ enterprise clients, shipping 35+ features a month at ≥95% test coverage. By night, I build autonomous AI pipelines, observability platforms, and a production e-commerce store solo — proof the ideas work end to end, not just in a demo.

Projects

• GitaFlow — Autonomous AI Content Generation Engine (2025 — Present). An autonomous pipeline that turns each of the 701 Bhagavad Gita verses into a narrated, captioned video — RAG-grounded scripts, locally-run voice-clone narration, local diffusion scene art, and local Whisper captions, orchestrated through a 7-stage agent pipeline on a swappable multi-LLM layer. Stack: Python, SQLAlchemy, Alembic, Gemini/Claude/GPT, LlamaIndex, ChromaDB, faster-whisper, Diffusion, Streamlit, APScheduler. Outcome: 144 commits driven test-first; zero repetitive output across hundreds of generated videos via dedup-store logic.
• Trulla — AI-Native Engineering Platform (Nov 2021 — Present). A multi-tenant pharmaceutical supply-chain SaaS for the U.S. 340B drug-pricing compliance program serving 30+ enterprise clients, plus the in-house agentic engineering platform — 12 AI rules, 14 agent skills, 8 MCP servers, and an LLM Council — built on top of it. Stack: Django 6, DRF, PostgreSQL, Celery, Redis, AWS Lambda, ECS Fargate, CloudFormation, MCP, LangChain, LangGraph. Outcome: 35+ production tickets/month at ≥95% pytest coverage across 30+ tenants; zero high-risk architectural failures in 12 months.
• LLM Council — Multi-Agent Decision System (2024 — Present). A 5-advisor multi-agent decision pipeline — independent analysis, anonymous peer review, then synthesis — used for high-stakes architecture, migration-safety, and refactor-scope calls on a regulated production codebase. Stack: Claude (Anthropic), GPT (OpenAI), LangGraph, MCP-aware context injection. Outcome: Zero high-risk architectural failures in 12 months of high-stakes calls.
• Agentic SRE — Observability Platform (2025 — Present). An end-to-end agentic AI observability platform built from a blank repo — anomaly detection, incident correlation, a 4-agent Gemini triage pipeline grounded by RAG over runbooks, and a React analytics dashboard. Stack: Python, FastAPI, SQLAlchemy 2.x, Pydantic v2, Google Gemini, RAG, React 18, Vite, TypeScript, Docker, GitHub Actions. Outcome: 98.75% test coverage across 246 tests; review score raised to 9/10 after closing 8 findings test-first.
• HR Comp Analytics — Compensation Intelligence Platform (2025 — Present). A compensation-analytics tool for a simulated 10,000-employee org, computing compa-ratio, range penetration, and percentile pay outliers via SQL window functions, with a React/TanStack dashboard. Stack: FastAPI, SQLAlchemy 2.x, Pydantic v2, SQLite, React 19, TanStack Query/Table, Recharts, Tailwind, shadcn/ui, Fly.io, Vercel. Outcome: 99% test coverage across 69 tests; 10,000 records seeded in 0.09s.
• Label by Mahi — Production E-Commerce Platform (Live in production). A live e-commerce platform — catalog, cart, checkout, custom coupon engine, wishlist — with real Razorpay payments, personally deployed and operated on AWS EC2 and Oracle Cloud. Stack: Django 5, PostgreSQL, django-allauth, HTMX, Razorpay, Nginx, Gunicorn, systemd, Certbot, AWS EC2, Oracle Cloud. Outcome: Live production store processing real payments across 7 Django apps, solo-operated end to end.
• Fraud Detection — AI Biometric Verification (2022 — 2023). A facial-recognition and behavioral-analytics fraud-detection module integrated into Trulla for identity verification during transaction onboarding. Stack: Python, OpenCV, Machine Learning, AWS Lambda, PostgreSQL. Outcome: 30% improvement in fraud-detection rate; $500K annual reduction in losses.

Skills

• AI / LLM: LLMs (Claude · GPT-5 · Gemini), LangChain, LangGraph, LlamaIndex, ChromaDB, Agentic AI, Prompt / Context Engineering, RAG, MCP, Multi-Agent Systems
• Languages: Python, JavaScript, TypeScript, SQL, HTML5, CSS3
• Backend: Django 6, DRF, Flask, FastAPI, SQLAlchemy, Pydantic
• Frontend: React 18/19, Vite, Tailwind CSS, shadcn/ui, TanStack Query/Table, Recharts, HTMX, Streamlit
• Databases: PostgreSQL (schema-per-tenant), MySQL, SQLite, Redis, Elasticsearch, Firebase, Alembic
• Cloud / DevOps: AWS (Lambda · ECS · CFN · CloudWatch · SQS/SNS · EventBridge), Azure, Oracle Cloud, Fly.io, Vercel, Docker, CI/CD, Nginx/Gunicorn
• Testing / Quality: pytest, TDD (≥95% coverage), Selenium, Ruff, Black, Mypy, pip-audit
• Domain: Healthcare / 340B, HL7, EDI 810/832/856, HIPAA-aware, Fintech, E-commerce

Experience

• Senior AI Engineer at Nagarro (Apr 2023 — Present) — Architected the AI-native engineering workflow shipping 35+ features/month on a regulated healthcare SaaS.
• Senior Python Developer at Nagarro (Nov 2021 — Mar 2023) — Backend / data engineering: ETL pipelines, multi-tenancy, fraud detection.
• Python Developer at Skill-Up Technologies (May 2019 — Nov 2021) — eLearning chatbots, NLP, Microsoft Graph integrations.
• Python Developer at Zingmobile (Sep 2018 — May 2019) — Elasticsearch optimization, web scraping at scale.
• Python Developer at AppAmplify (Jul 2017 — Sep 2018) — Batch automation, ad-tech tooling.

Contact: samit.pawar7711@gmail.com · GitHub · LinkedIn