Your dev profile
backed by code.

Collect evidence with the CLI, build your profile with your coding agent, and publish when it is ready.

$ npm install topsignal
View on GitHub

You've been making do

Your deserves better.

  • LinkedIn doesn't speak dev

    Skills endorsed by people who've never seen your code. A profile that can't show what you actually shipped.

  • GitHub isn't a profile

    A contribution graph is not a portfolio. There's no way to tell your story, control your narrative, or showcase your work.

  • You're being searched

    Recruiters are already Googling you. Share your best code stories and outcomes.

www.topsignal.dev
About karpathy

Overview

Andrej is a world-class deep learning researcher and engineer with 16 years of GitHub history and a focus on building minimal, educational implementations of large language models (LLMs) and neural networks. He is best known for creating highly starred open-source projects like nanoGPT and llm.c, which are widely used across the AI community, signaling exceptional technical depth and the rare ability to make complex concepts accessible. With over 151,000 followers and multiple repositories exceeding 10,000 stars, he is one of the most influential AI practitioners on GitHub today.

Role Fit

AI/ML Research EngineerDeep Learning Educator / Developer AdvocateAI Infrastructure EngineerLLM Systems Architect
Strengths

Elite LLM & Deep Learning Expertise

Built nanoGPT (55k+ stars), minGPT (24k+ stars), llm.c (29k+ stars), and nanochat (50k+ stars) — all focused on training and fine-tuning large language models from first principles, demonstrating mastery of the full LLM stack.

Low-Level Systems Programming

llm.c implements LLM training in raw C and CUDA (15.8% of codebase is C, 0.6% CUDA), showing the ability to go far below Python abstractions and optimize at the hardware level — a rare skill in ML.

Exceptional Ability to Simplify Complexity

Repositories like micrograd, minGPT, and nn-zero-to-hero are explicitly designed to teach neural network concepts from scratch, with millions of learners using them globally, indicating elite-level pedagogical and communication skills.

High Community Impact & Thought Leadership

151,365 GitHub followers, multiple repos with 10k–55k stars, and a contribution to the official Python CPython repository signal that his work sets standards others follow.

Prolific & Consistent Output

63 public repositories spanning GPT architectures, tokenization (minbpe), inference engines (llama2.c), and research tooling (arxiv-sanity-preserver), demonstrating sustained, high-quality output across many years.

Standout Projects
nanoGPT

The most starred repo at 55k+ stars, it is the go-to reference implementation for training medium-sized GPT models. Its widespread adoption in academia and industry proves both technical excellence and real-world utility.

llm.c

Implementing LLM training in raw C and CUDA with 29k+ stars is an extraordinary technical feat that demonstrates systems-level mastery most ML engineers lack, and it directly influences how practitioners think about performance optimization.

nn-zero-to-hero

A Jupyter Notebook-based course with 21k+ stars that has become a canonical resource for learning neural networks from scratch, highlighting his unique ability to bridge research and education at scale.

micrograd

A tiny autograd engine with 15k+ stars that replicates PyTorch's core functionality in minimal code — widely regarded as one of the best teaching tools in deep learning, showcasing elegant, purposeful engineering.

Built differently

Your profile. Your rules.

  • Open source, all the way down

    TopSignal is fully open source. No black boxes, no vendor lock-in. Inspect it, contribute to it, or self-host.

  • You control the evidence

    The CLI saves private evidence for your coding agent to review. Choose which repositories to include and inspect your completed profile before publishing.

  • Make it yours

    Customize your profile. Pick your theme, highlight what matters. This is your page — not a template built for an ATS.

  • Share with recruiters, on your terms

    When you're ready, generate a clean shareable link that shows exactly what recruiters should see — skills verified by real commits, not self-reported.

  • Show off your skills

    Let your work speak for itself. Share your profile with teammates or your network and let real commit history back up your reputation.

  • See where your gaps are

    Discover skills you're close to mastering and areas worth investing in. Honest insights from your actual work history — no guesswork.

How it works

From your code to your profile.

  1. 01

    Collect your evidence

    Point the CLI at GitHub and your local code. It writes your profile and sampled evidence to files for your coding agent.

    npx topsignal collect johndoe --dir ~/Code --output profile.json --evidence evidence.json
  2. 02

    Ask your coding agent

    Your agent reviews the private evidence and writes its analysis. Attach that analysis, then review the completed profile.

    npx topsignal attach-analysis profile.json analysis.json --output completed-profile.json
  3. 03

    Publish & share

    Log in with GitHub, then publish the profile you reviewed to get a shareable link.

    npx topsignal publish completed-profile.json
View on GitHub

CLI FIRST

The terminal is where agents live.

The CLI collects structured data. Your coding agent supplies the analysis through the tools and model you already use. Review the profile, then publish it with the CLI.

topsignal — bash
$

Open source at the core

Built by developers, for developers.

Your profile is a portable JSON file. The CLI collects the evidence, your coding agent adds the analysis, and you decide what gets published. Keep the source evidence in your own workflow.

  • Your data is open

    Your developer profile data is yours — open, portable, and never locked behind a platform. No proprietary formats, no black boxes.

  • Your data, your export

    Download everything, anytime. No hostage data, no lock-in. Take your profile wherever you go.

  • Your agent, your choice

    Use your coding agent to analyze the collected evidence. TopSignal runs no model and publishes only the profile JSON you choose.