Achyuth Madabhushi

Musings, forkings, readings

Here lies a smattering list of things I'm interested in and digging into. It's by no means exhaustive or authorative, it's just a smattering.

This fork list is honest about what I'm actually into in a way that a résumé can't be.

AI & ML engineering

The most recent additions are the Anthropic Python SDK and a Go-based educational RAG (ai-rag-with-go using Temporal MinIO and Ollama, an end-to-end ETL, embeddings and semantic-search). Older clones I keep around as references: PyTorch, Detectron2 (object detection / segmentation), and OpenAI's Spinning Up deep-RL curriculum. MLQM is the curiosity dabble — applying ML to quantum systems.

Climate & energy

PowerGenome generates inputs for power-systems models — useful for thinking about decarbonization scenarios. contrastive_sensor_fusion implements unsupervised representation learning for remote sensing; the ability to learn from unlabeled satellite/sensor data is increasingly load-bearing for climate work. Both reflect the part of my background that's still tracking the energy transition. Remote sensing and GIS were part of my first foray into information systems dating back to 2015. In my first year of graduate school, one of my final projects was using LandSat imagery for a raster analysis of Houston, post Hurricane Harvey.

Cybersecurity

GOAD (Game of Active Directory) is a self-contained AD pentest lab I use to keep offensive-security muscles warm. Google's security-research repo is where I read real vulnerability writeups. The OSCP cheat sheet is exactly what it sounds like, and security_learning_scripts (my own) is the messy notebook behind it.

Responsible AI

DSSG's Aequitas — a bias and fairness audit toolkit. I think about this more often than I write about it. I also recommend the HELM framework coming out of Stanford's CRFM. I know there's mixed opinions there but for heuristically comparing one LLM to another, it's excellent. I don't think any metric exists to truly measure intelligence, not even IQ. I like to think of intelligence like the color wheel, where each color serves a different purpose, and everyone has their favorite.

The through-line, if you need one: data and software that interact with the messy, socially consequential parts of the world — climate, security, fairness — and the engineering plumbing that makes them actually work. I've had the gracious experience of being around several amazing mentors, who have inspired me to find the intersections of my interests and explore.

Reading list

The rolling list — the textbooks I keep open in tabs and intermittently make progress on:

  1. Intro to Cryptography (Cornell CS 4830 lecture notes)
  2. Intro to Control Theory

Books I've read and recommend

Nonfiction

Philosophy & Psychology

Science Fiction

Systems and Design

Comedy