I'm Harshit, an AI Engineer at GSK working at the intersection of generative AI, platform engineering, and enterprise delivery. I build production-grade RAG applications, multi-agent systems, and reusable LLM tooling that make complex, regulated workflows simpler and faster.
My sweet spot is the full journey: understanding the problem, shaping the architecture, building the AI and application layers, and shipping with evaluation, observability, security, and CI/CD in place.
- Agentic AI: orchestration, memory, tool use, guardrails, and human feedback with LangGraph and LangChain
- Retrieval systems: vector and graph RAG, document intelligence, knowledge graphs, reranking, and evaluation
- AI platforms: reusable Python libraries, FastAPI services, data pipelines, CI/CD, tracing, and cost monitoring
- Responsible delivery: secure cloud architecture, access control, redaction, quality monitoring, and GxP-aware engineering
| Project | What it does | Built with |
|---|---|---|
| LangHost | Open-source, self-hosted LangGraph Agent Server with production Postgres and Redis | Python · LangGraph · Docker |
| GitVideo | Turns a GitHub repository into a polished AI-generated showcase video | Python · Generative AI |
| OCR Compare | Compares PDF-to-Markdown OCR pipelines for context-engineering workflows | Python · Streamlit · OCR |
| Gemini Data Extractor | Extracts structured document data against a user-defined schema | Python · Gemini · Streamlit |
- 🏆 Winner of the GSK × Databricks Hackathon 2025
- 🎓 Pursuing an MSc in Artificial Intelligence at the University of Leeds
- 🌱 Exploring better evaluation, context engineering, and reliable agent architectures
- 🤝 Always happy to talk about applied AI, open source, or an interesting problem
Build useful things. Measure what matters. Keep learning.