Skip to content
View nikhilcherry's full-sized avatar

Block or report nikhilcherry

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
nikhilcherry/README.md

I build things end to end: ML pipelines, hardware and firmware, mobile apps, and the tooling around them. Most of it started at a 24 to 48 hour hackathon, and what is below is what I would still show someone. I like work where the number is measured (held-out, calibrated, reproducible) and where the README says what the project does not do.

Selected work

assay replay page

assay · fraud investigation on a graph
An agent that investigates a flagged transaction on TigerGraph and files a case with the query behind every claim. Held-out AUC 0.964 vs the bank's 0.866; no LLM on the decision path. Live replay.

Chaukanna

chaukanna · live scam-call guardian
Whispers a warning only the victim hears, an AI agent cuts the scammer off, n8n files the evidence. Caught 28 of 28 scam calls with 0 false alarms on 19 ordinary ones. Demo.

Four forward models fitted to one light curve

arvyo · exoplanet or impostor?
A star dims. Planet, eclipsing binary, blend or starspot? Fits all four from physics on TESS/Kepler data and ranks them by BIC. Found a real near-tie between planet and blend and documented it.

Connect architecture

connect · reach a car's owner, no numbers shared
QR tag on the windshield, private two-way chat, works offline through an acoustic modem I wrote in Dart. 139 backend checks, 146 app tests, ₹0/month to run. Top 7 at the iQOO grand finals.

jitter blaming a timezone flake

jitter · why a test is flaky
Shakes twelve sources of nondeterminism, finds the one that matters statistically, and hands back a repro that fails 10 out of 10. Zero dependencies.

strata report on Flask

strata · code archaeology
Crosses git churn with code shape to show where a repo's risk lives, including files that change together with no import between them. One HTML file, zero dependencies.

Watch the demos: assay, 3 min · chaukanna, 2 min · connect, reel · arvyo, 40 s · jitter, 40 s · strata, 40 s

Developer tools

Small CLIs with tests and CI, each doing one thing:

vein what your code actually runs: execution diffs and runtime dead code
carve shrink a failing repo to the least that still fails the same way
replayr record any HTTP/SSE/WebSocket backend once, replay it forever
hoist a local app on a public HTTPS URL with one command
roam a Blender export to a walkable multiplayer web page in 172 KB
kinetix beat-synced vertical reels from code, shipped as a Claude Code skill
More: hardware, sensing and on-device AI
  • chowki: a delay-tolerant Bluetooth mesh for mountain trail booths, with a local Gemma "brain" (chowki-brain). Built at the Google DeepMind Bangalore hackathon.
  • prism: a laptop's speaker and microphone as science instruments: ultrasonic room geometry (prism-echo), breathing detection (prism-breath) and everything exposed as MCP tools (prism-mcp).
  • DriveLink: a V2V communication layer for cars, with an ML engine, an isometric simulator and the website.
  • whosaid: on-device, speaker-attributed captions for deaf and hard-of-hearing users, built from small browser toolkits (ASR, VAD, face tracking, caption HUD).
  • hk-voice-io: local offline speech toolkit (VAD, Whisper, Piper) that powers my local dictation tool.

Hackathons

Result Event
🥇 1st, ₹50,000 MSRIT Pariyavaran
🥇 1st, $350 ADHD Hacks
🏅 AIR 5 IIT Delhi Blueprint
4th HackSurgeX
🏅 Top 7 iQOO hackathon, grand finals (connect); top 10 in the June round
Top 10 Google DeepMind hackathon, Bangalore

Also shipped at Hacker House Goa 2026 (assay) and Hack Sprint 2026, MIT Bengaluru (chaukanna).

How I work

  • Measure first. Held-out splits, calibration tables, 10-out-of-10 replays. If a number is an assumption, the README says so.
  • Write down the limits. Every project above has an "honest limits" section, because an interviewer finds them anyway.
  • Ship the demo, then the tests. The hackathon builds are demo-first; the dev tools and the harder projects carry tests that run in CI.
Resume and selected projects: nikhil-av.vercel.app

Pinned Loading

  1. arvyo arvyo Public

    A star dims. Something crossed it. Arvyo works out what — planet, eclipsing binary, blend or starspot — by fitting all four from physics. 11 repositories, one frozen data contract.

  2. assay assay Public

    Agentic fraud investigator on TigerGraph with calibrated, measured probabilities. Hacker House Goa 2026, Task 4.

    Python

  3. chaukanna chaukanna Public

    Live scam-call guardian: whispers a warning only the victim hears, AI agent cuts the scammer off, n8n files the evidence. Sarvam + Jev + LiveKit + n8n. Hack Sprint 2026.

    Python

  4. jitter jitter Public

    Find out why a test is flaky, not just that it is. Twelve isolated sources of nondeterminism, statistical attribution, and a repro that fails 10 out of 10.

    Python

  5. strata strata Public

    Code archaeology: crosses git history with code structure to show where the risk in a repository actually lives. One self-contained HTML report, zero dependencies.

    Python

  6. connect connect Public

    Reach a car's owner about their car without either side sharing a phone number. QR tag, private two-way alerts, on-device ML. Flutter + Supabase.

    Dart 1