Prepared for Riot GamesBase version
Domain models for a live competitive game: patch-aware retrieval, agentic harnesses, and classifiers that production staff actually rely on.
Top 25
Riot x AWS hackathon, globally
25+
custom playstyle classifiers
2 wks
to earn official Riot data API access
About me
I am Vansh Bordia, an AI engineer studying Data Science and AI at IIT Madras. I read the diffusion papers before they were mainstream, then built on top of them: domain-tuned models, RAG over live data, agentic harnesses serving real users.
My bias is production over papers. Models that do not survive contact with real data and real latency budgets do not interest me; systems that coaches and analysts rely on daily do.
Why me
I read the diffusion papers before they were mainstream, then built on top of them: a VALORANT-specific Gemma layer, RAG over live data, agentic harnesses routing tools and models for real coaching staffs.
I build production AI: a VALORANT-specific Gemma layer with RAG over live data, agentic harnesses routing tools and models, and ML pipelines that evolve with incoming matches. I read the diffusion papers before they were mainstream, then built on top of them.
VALORANT-specific Gemma models plus RAG over live patch data
Agentic harness routing tools, models and queries; sub-500ms P90
Top 25 finish globally in the Riot x AWS hackathon
Role stack
Python / PyTorch / LLM Fine-Tuning / RAG Pipelines / Agentic Systems / Vector Databases / Feature Engineering / Model Evaluation / MLflow / Scikit-Learn / Prompt Engineering / MLOps
Relevant work
Esports intelligence platform, acquired by a pro org
Type-safe Python library for Valorant data, published on PyPI
Interactive SQL schema visualizer, 100% client-side
Contact
Need AI that ships to prod? Talk to me.