Skip to content
View Hetul803's full-sized avatar

Achievements

Achievement: Pair ExtraordinaireAchievement: Pull Sharkx2Achievement: QuickdrawAchievement: YOLO

Achievements

Achievement: Pair ExtraordinaireAchievement: Pull Sharkx2Achievement: QuickdrawAchievement: YOLO

Block or report Hetul803

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
Hetul803/README.md

Hi, I'm Hetul 👋

I'm a CS graduate student and builder focused on AI infrastructure, vector retrieval systems, and autonomous agent tooling.

My work spans performance-critical backend systems, large-scale experimentation pipelines, and governance tooling for AI agents.

🔧 What I'm building

Aegisure — AI Agent Governance & Audit Platform

A vendor-neutral control and audit plane for AI coding agents (Claude Code, Codex, Cursor, Copilot).

  • Intercepts and audits agent actions at the commit/PR layer
  • CLI-first, integrates with existing GitHub workflows via webhooks
  • Targets SOC2 compliance automation and enterprise AI governance
  • Published on PyPI (v0.2.1)

➡️ PyPI · GitHub

VecLite — Vector Database Engine (MCN Algorithm)

A production-grade vector database built on Memory Compression Networks — achieving massive compression without sacrificing recall.

  • 12.71× compression with exact recall parity vs FAISS (Recall@10 = 39.34% — identical to FAISS IndexFlatIP)
  • Outperforms FAISS on fraud detection recall (4.30% vs 3.55%)
  • Non-blocking index finalization via atomic snapshot swap — zero downtime during 137s cluster rebuilds
  • 0 errors across 23,269 concurrent queries during online mutation test
  • Multi-tenant SaaS backend: 5 plan tiers, token-bucket rate limiting, no external database

➡️ Live demo: veclite.com · Backend (open source)

GSIN — Global Strategy Intelligence Network

A self-evolving algorithmic trading research platform for systematic strategy discovery.

  • Evaluated 45,672 strategies via genetic algorithm evolution across 9 users
  • Walk-forward backtesting + Monte Carlo simulation + regime-aware scoring
  • Discovered and diagnosed a WFA trade aggregation bug producing systematically inflated win rates across the entire strategy corpus
  • Multi-objective scoring: Sharpe, win rate, drawdown, out-of-sample stability
  • Full-stack: Next.js 14, FastAPI, Supabase, Railway

➡️ Live demo: gsin.trade

🧠 Research

Hyperspectral Imaging + ML — Auburn University at Montgomery
Targeting publication in Atmospheric Measurement Techniques (AMT)
Single-camera design hypothesis confirmed via radiance transfer experiment. Key ML results complete.

🛠️ Tech stack

Languages: Python, TypeScript, SQL
AI/ML: Vector search, genetic algorithms, walk-forward backtesting, scikit-learn, NumPy, Hugging Face
Backend: FastAPI, Next.js, REST APIs, WebSockets
Infrastructure: Supabase, Railway, Vercel, Docker
Tools: Git, PyPI packaging, GitHub Actions

📫 Connect

Pinned Loading

  1. Hetul803 Hetul803 Public

    MSCS @ AUM | Quantitative Research in training | Python, C++, Trading Systems, NLP

    1

  2. Veclite_backend Veclite_backend Public

    Python

  3. Veclite_frontend Veclite_frontend Public

    Compressed vector storage with exact recall to save vector DB cost

    Python