Skip to content
View akshayb7's full-sized avatar

Block or report akshayb7

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
akshayb7/README.md

Akshay Bhardwaj

I build the scaffolding that makes AI agents reliable enough to actually ship — and I lead the org that ships it.

Director of Data Science & AI at Mechademy. I walked in as the company's first data scientist; today I run a ~70-person org with a ~20-person DS/AI/ML core at the center of it.

How I got here, short version: petroleum engineer → watched AlphaGo play Move 37 → took it personally → taught myself into AI and never looked back. Ten years on I'm still chasing the same thing — go deep first, orchestrate second. Depth earns the right to orchestrate.

Most of what I build now is agent harnessing: the orchestration, evals, and guardrails that turn a clever token-predictor into something you can put in front of real users. The model was never the point. Deployment was.

Stuff I've built because I couldn't not

  • Ouroboros — point it at an idea, get back a tested, merge-ready PR. Built around the failure modes that quietly wreck agent teams.
  • Transformer, from scratchAttention Is All You Need in raw PyTorch. Built to be derived, not imported.
  • faceless-ai — folklore in, finished horror short out. Five model modalities, one human gate.
  • Raphael — a voice assistant for my second brain that never leaves my laptop. Fully local, sealed loop.

Toolbelt — Python · LangGraph · Ray · PyTorch · Dagster · MLflow · Claude / Gemini / GPT · Whisper.cpp · Ollama · FastAPI · TimescaleDB · Iceberg · AWS

I write, too — long-form, first-principles, occasionally funny. An 8-part series on Ray, and the one where I argue attention is a potluck.

Portfolio · Blog · LinkedIn

Pinned Loading

  1. podcast-summarizer podcast-summarizer Public

    Python

  2. equity-analyzer equity-analyzer Public

    Python

  3. notion-knowledge-distiller-mcp notion-knowledge-distiller-mcp Public

    Python

  4. ai-blog ai-blog Public

    HTML