Skip to content
View AravindKurapati's full-sized avatar
3 followers · 0 following

Achievements

Achievement: Pull Sharkx3Achievement: Pair Extraordinairex3Achievement: YOLO

Achievements

Achievement: Pull Sharkx3Achievement: Pair Extraordinairex3Achievement: YOLO

Block or report AravindKurapati

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
AravindKurapati/README.md

hey, i'm aravind

i work on ML and AI systems and make the most of free cloud credits
(shoutout to Modal for keeping me GPU-rich while unemployed)

recent stuff

  • built a flight recorder for Claude Code & codex agents.....captures every tool call, shell command, error and API cost per session

  • built a CLI to track Claude session windows and weekly usage after realizing I was burning 10-15% of my limit on sessions I never finished. also added a tool that indexes them, embeds them with MiniLM, and whispers "you already tried this" before I start something redundant

  • built a retro terminal dashboard that uses an agentic loop to scan GitHub for creative coding projects across NYC, London, SF & LA.

  • benchmarked LLaMA 3.1 8B vs Mistral 7B on SEC 10-K filings on Modal A10G. Mistral is 4.5x faster on TTFT. built the full RAG pipeline: vLLM + FAISS + FastAPI + Streamlit

  • spotted a repo tracking leaked system prompts from major AI labs on Twitter. forked it and built a dashboard on top so people can actually see the diffs and changes over time

  • ran a full scale inference of AlphaFold2 (DeepMind) on GCP. hit every CUDA error imaginable, debugged all of them, documented the whole thing

  • predicted cancer subtypes from pathology slides at NYU Langone using Self Supervised Learning + attention MIL on gigapixel WSIs

  • published 2 research papers in 2021 before the AI boom, before Claude/ChatGPT, before any of this was cool! benchmarked projects written from scratch and debugged entirely on Stack Overflow like a normal person

currently: open to AI/ML & engineering. FDE & relevant roles too! not currently: employed (hence the aggressive committing)

tools & infrastructure i've used

(used their products/infra.... not claiming i work in every single company, recruiters!)

arvind.kurapati@gmail.com · linkedin medium

Pinned Loading

  1. claude-burnrate claude-burnrate Public

    Track your Claude session windows and weekly usage. Never waste limits again.

    Python 1

  2. claude-best-practices claude-best-practices Public

    A living knowledge base of Claude, Claude Code, skills, MCP, and agentic workflow best practices collected from practitioners and kept updated.

    5

  3. alphafold-prediction alphafold-prediction Public

    Full AlphaFold2 inference on GCP. NVIDIA driver debugging, Docker setup, 2.8TB database downloads, and insulin structure prediction

    1

  4. llm-serving-sec-filings llm-serving-sec-filings Public

    RAG over SEC 10-K filings: LLaMA 3.1 8B vs Mistral 7B benchmarked on Modal A10G. vLLM + FAISS + FastAPI + Streamlit.

    Python 1

  5. TCGA-Lung-Inception-NYU-HPC TCGA-Lung-Inception-NYU-HPC Public

    Classifying lung cancer subtypes (LUAD vs LUSC) from TCGA histopathology slides using InceptionV3 on NYU HPC. Built on top of Barlow Twins SSL embeddings and the HPL framework.

    Python 1

  6. research-knowledge-vault research-knowledge-vault Public

    Living research knowledge system for computational pathology and biomedical ML - built with arscontexta, covering FL subtype prediction, TCGA lung cancer classification, and AlphaFold structure pre…

    Shell 1