I build at the intersection of applied AI, quantitative systems, scientific computing, and product engineering. Based in Bengaluru.
zeroH — A Python layer for grounded memory, retrieval, claim verification, and abstention. Wraps any LLM with a deterministic verify-reask-abstain loop using SQLite-backed memory, sentence-aware chunking, and confidence scoring. No heavyweight ML dependencies.
ZANE — AI-native pharmaceutical OS unifying target intelligence, molecule generation, free-energy physics, preclinical safety triage, and lab execution. Funded by AMD and NVIDIA.
vecomp — A Rust vector index with Python bindings that shrinks high-dimensional embeddings to ~2GB via quantization-aware retrieval.
Tetra Quant — Quantitative trading research infrastructure on OpenBB. Data ingestion, backtesting, and signal analysis.
Concepts — AI-native note-taking app applying grounded retrieval to personal knowledge management.
Hydra Fund — Software and research infrastructure for an AI-native investment platform.
Python, TypeScript, Rust, Swift, Go, SQL — React, Next.js, FastAPI, OpenBB — PostgreSQL, SQLite, Docker, AWS
Determinism over heuristics. Stochasticity is a tool, not a default.
Composable over monolithic. Small, standard-library-first primitives that compose well.
Evaluate honestly. Grounding verification, abstention rates, and failure mode analysis over inflated metrics.
Ship real things. Applied research that produces software people can use.
© 2026 · cosmic-hydra