Skip to content
View shipitdev's full-sized avatar
13 followers · 13 following

Achievements

Achievement: Pull Sharkx2Achievement: Pair Extraordinaire

Achievements

Achievement: Pull Sharkx2Achievement: Pair Extraordinaire

Block or report shipitdev

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.

Content in all repositories owned by your account will be closed.
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
shipitdev/Readme.md

Harsh K. Singh

Software Engineer | Systems & Applied ML

confidence: high
also: i tested it three times

I architect distributed systems that deliver AI hallucinations with sub-millisecond latency.

Shipped

  • LambdaSearch: A Learning-to-Rank search engine built with XGBoost LambdaMART and Elasticsearch. Trains on synthetic click data to re-rank BM25 results, evaluated by NDCG@10. (Because standard full-text search just wasn't complicated enough).
  • Hackawon: A static, serverless platform that aggregates hackathons from Devfolio, Unstop, and MLH, then generates project ideas grounded in past winning submissions. Runs entirely on scheduled GitHub Actions: no backend, near-zero operating cost. (Turns out you don't need a server if you're stubborn enough about GitHub Actions).
  • Halo: A macOS overlay app. !think open-source Cluely/ParakeetAI: captures screen, mic, and meeting context in real time and streams live AI assistance across OpenAI, Gemini, and Anthropic models. (An always-on-top window that knows what you're doing. Slightly unsettling, very useful though).

Open Source

Currently working on my Open-source contributions. I'm currently contributing across Flatcar (test automation harness for sysext images) and CNCF projects, also exploring agent-based modeling. Past contributions include NumPy, scikit-learn, statmodels and a few more around the PyData ecosystem.

Current AI Stack (subject to change every time a new model drops)

LLMs:        whatever has a good API and doesn't rate-limit me at 2am
Retrieval:   pgvector until it hurts, then actually learn the fancy stuff
Agents/MCP:  build the tool first, trust the model second
Infra:       my own PC, until it can't take it anymore

Tech Stack:

C C++ Python JavaScript Go PHP FastAPI NodeJS Express.js NumPy Pandas Matplotlib PyTorch TensorFlow scikit-learn Docker Linux Server Administration Elasticsearch Redis GitHub Actions

Profile Views

Pinned Loading

  1. cooked cooked Public

    An LLM-powered roast engine that analyzes your Spotify listening data and tells you exactly what it says about you. Brutally

    JavaScript

  2. halo halo Public

    An interview copilot overlay, quietly helping you think while pretending you’re just really good at interviews.

    JavaScript 1

  3. neural-scratch neural-scratch Public

    A tiny neural network built from scratch because apparently the best way to understand deep learning is to wrestle with gradients manually and pretend you’re not one bug away from a full existentia…

    Python 1

  4. LambdaSearch LambdaSearch Public

    Learning-to-Rank search engine using XGBoost LambdaMART and Elasticsearch. Trains on synthetic click behavior to re-rank BM25 results, evaluated by NDCG@10 on a held-out query set.

    Python

  5. cloud-metadata-simulator cloud-metadata-simulator Public

    A simple Go server that emulates cloud provider instance metadata services (IMDS) and serves #cloud-config user-data over HTTP.

    Go

  6. hackawon hackawon Public

    Hackawon: discover hackathons and their problem statements, learn from past winners, and cook up your next winning idea before the deadline cooks you.

    Python