B.Tech AI · G.H. Raisoni University (2024–2028)
Microsoft Learn Student Ambassador — top 0.01% globally
Nagpur, India · open to research internships
I build ML systems that are meant to be understood, not just run.
My projects focus on evaluation and reliability — the parts of the ML stack that most tutorials skip. How do you know your RAG pipeline is actually retrieving the right chunks? How do you catch hallucinations before they reach users? How do you validate that your NLP preprocessing is actually helping downstream accuracy?
Recent work
LLM Output Validation & RAG Framework
Automated hallucination detection and retrieval validation for LLM pipelines.
94% retrieval accuracy · 12,000+ records/day · LangChain · Pinecone · Anthropic API · FastAPI · AWS
End-to-End NLP Data Pipeline
BERT-based anomaly detection and schema validation with real-time monitoring dashboard.
93% downstream accuracy · 20,000+ records/day · BERT · MongoDB · React · D3.js
Stack
core = ["python", "pytorch", "bert", "transformers", "langchain", "spacy"]
infra = ["fastapi", "docker", "aws", "postgresql", "mongodb", "git"]
research = ["experimental design", "ablation study", "reproducible research"]
apis = ["anthropic", "openai", "gemini", "azure ai", "pinecone"]Elsewhere
- Microsoft: designing AI programs for 400+ participants as a Learn Student Ambassador
- GH Raisoni: founded Phoenix AI Club, 200+ members, #1 technical community year 1
- GSSoC: 120+ code reviews across 4 open-source ML projects
hemant_189@outlook.com · linkedin.com/in/hemantchilkuri