Skip to content

Hotragn

AI engineer building agents and graph-RAG systems that verify their own output.

Selected work

Projects

All work

Open source

Open-source contributions

  1. 95.9K stars
    8 merged

    vLLM (my merged pull requests on GitHub, opens in a new tab)

    One of the most widely used engines for running large AI models on servers. I fixed 8 bugs, mostly in how long prompts are trimmed and how uploaded images and videos are checked before the model sees them.

  2. 13.6K stars
    11 merged

    Hugging Face speech-to-speech (my merged pull requests on GitHub, opens in a new tab)

    Hugging Face's toolkit for building voice assistants from open models. I fixed how it detects which language someone is speaking, and kept private conversation transcripts out of the server logs.

  3. 1.1K stars
    8 merged

    nnsight (my merged pull requests on GitHub, opens in a new tab)

    A research library for looking inside AI models while they run. I fixed bugs, added support for Python 3.14, and made its automated tests cover every Python version it supports.

  4. 8.9K stars
    19 merged

    OpenBot by CopilotKit (my merged pull requests on GitHub, opens in a new tab)

    Open-source AI coworkers for teams. My changes make them safer and easier to audit: they refuse tools a team never added, answer only from documents the person asking may read, and record why a message was not delivered.

  5. 41.4K stars
    2 merged

    AirLLM (my merged pull requests on GitHub, opens in a new tab)

    Runs very large AI models on small graphics cards. I made it pick the best hardware on the machine instead of assuming an NVIDIA card, which closed three open bug reports.

  6. 165.6K stars
    18 agents · 8 fixes

    agency-agents (my merged pull requests on GitHub, opens in a new tab)

    One of GitHub's most-starred collections of ready-made AI agents. I wrote 18 of its specialist agents, from Statistician to FinOps Engineer, and fixed 8 bugs in its installer. The maintainer merged most of my agents in combined batches that credit me.

Experience

Where I've worked

About
  1. May 2026 - Present

    Open source

    Open-Source Contributor

    • 84 pull requests merged across 11 projects since graduating.
    • vLLM: fixed how prompts get cut to length and how image and video inputs are checked before they reach the model.
    • Hugging Face speech-to-speech: fixed spoken-language detection and stopped conversation transcripts from ending up in server logs.
    • CopilotKit OpenBot: made its agents refuse tools they were never granted and answer only from documents the user can see.
    • agency-agents: wrote 18 of its specialist agents and fixed bugs in its installer.
  2. Sep 2024 - May 2025

    Northeastern University

    Graduate Research Assistant

  3. Jan 2025 - Apr 2025

    Northeastern University

    Teaching Assistant

  4. Jan 2024 - May 2024

    Foruppo

    Software Engineering Fellow

  5. Feb 2023 - Jul 2023

    AppsTek Corp

    Research Intern

Writing

Published writing

All writing
  1. HackerNoon

    The OpenAI-Hugging Face Incident Was an Identity Failure Before It Was an AI Failure (on HackerNoon, opens in a new tab)5 min read

    OpenAI's agent escaped containment and hit Hugging Face. The fix isn't smarter models. It's the identity layer the Agentic AI Foundation was built to ship.

  2. HackerNoon

    MCP Security in 2026: Lessons From Real Exploits and Early Breaches (on HackerNoon, opens in a new tab)7 min read

    MCP became the standard way to wire tools into AI in 2026. GitHub token leaks, remote-code-execution bugs and supply-chain attacks show the risks that came with it.

  3. DEV

    My Second Dance with Hacktoberfest (2026) (on DEV, opens in a new tab)5 min read

    It all started with a sticker. Two years ago, I saw an MLH sticker on a friend's laptop 💻, looked up what it meant, and happened to find Hacktoberfest 2024 running at the same time.

  4. DEV

    Paper Ghosts: I screened 370 million old newspaper articles for ghosts. One was a magic lantern (on DEV, opens in a new tab)14 min read

    Why I built this. I've always wondered about ghost stories. American horror movies like The Conjuring say they're based on true events, and they felt real to me, so I wanted to see the real ones for myself.

Recognition

Recognition

  • Knowledge Graph Conference

    Poster-session winner as lead author, for TravelGraph-AI

    2026

  • AI4S 2026 (Springer)

    Best Paper Award as lead author, for GreenTrust-Crisis

    2026

  • Copyright Office, Government of India

    Copyright L-150423/2024 for ICAKS

    2024

  • Oracle Cloud Infrastructure

    Generative AI Certified Professional

    2024

  • IJRPR, vol. 4 no. 12

    Risk management in information security, cybercrime and cyber-stalking

    2023

    Read more
  • IRJET, vol. 10 no. 11

    Data augmentation for low-resource language fine-tuning of LLMs

    2023

  • Zenodo

    Open dataset for predicting job roles from skills and experience

    2023

Toolkit

Tools I use

  • Python
  • SQL
  • TypeScript
  • PyTorch
  • TensorFlow
  • Hugging Face
  • Ollama
  • MLflow / LangSmith
  • LangChain / LangGraph
  • Model Context Protocol
  • Google ADK
  • Neo4j / Cypher
  • Snowflake
  • PostgreSQL
  • ChromaDB
  • FastAPI
  • Streamlit
  • React / Next.js
  • Docker
  • GitHub Actions