chinmay.

CHINMAY SUJITH NAIR / COMPUTER SCIENCE

Human questions. Engineered answers.

I’m Chinmay, a computer science student at the University of Michigan. I build across real-time systems, agentic AI, and applied machine learning—with people at the center.

Explore my work Get in touch

University of Michigan / B.S. Computer Science, expected 2028

01 / SELECTED WORK

A few things I’ve built.

Two explorations in making complex information useful.

MISSION 01 / PROXIPROMPT.AI

Actual ProxiPrompt mobile welcome screen
ACTUAL APP / LOCAL CAPTURE

NEARBY PEOPLE. BETTER SIGNAL.

01 / AGENTIC AI + DISTRIBUTED SYSTEMS

ProxiPrompt.ai

MHacks ’26

Local questions.
Human-grounded answers.

A real-time intelligence platform that checks fresh evidence first, asks nearby people only when needed, and answers with confidence and provenance—not model guesses.

231

TypeScript tests passed in this review135 shared-core + 96 orchestrator tests

  1. Check fresh evidence
  2. Ask nearby people
  3. Answer with provenance
  • React PWA
  • TypeScript
  • Python uAgents
  • SpacetimeDB
  • Zod + Pydantic
  • Web Push
View source (opens in a new tab) Hackathon submission (opens in a new tab)
Explore the engineering

Built in 24 hours. My résumé-backed contributions span the Python uAgent, TypeScript orchestrator, SpacetimeDB state machine, and confidence engine.

FIVE IDEAS BEHIND THE SYSTEM

  • Grounded answersFresh evidence first. People fill the gaps.
  • Deterministic confidenceCode scores sources; the model plans and writes.
  • Privacy by ruleDatabase checks protect access and exact locations.
  • Safe recoveryIdempotent jobs and database timers handle retries.
  • One live workflowReact and Python share SpacetimeDB state.

Current source26 tables · 51 reducers · 29 views

Verification notePython: 189 passed; two Windows test-ID errors.

Project scopeTeam hackathon; not a production-readiness claim.

MHacks ’26 · Physical iPhone push/native geofencing not verified.

MISSION 02 / PATTERN RECOGNITION

Signal.
Not noise.

12 ECG LEADS4 DATASETS5 CLASSES

STYLIZED SIGNAL / NOT CLINICAL DATA

02 / APPLIED MACHINE LEARNING · 2025–2026

ECG independent research

Learning from
the heartbeat.

Independent research into 12-lead ECG arrhythmia classification, with an emphasis on what models learn—and how they generalize across datasets.

0.90

Peak weighted F1Across five clinically tailored classes

12 ECG leads

4 Datasets

5 Clinical classes

  • Python
  • ECG classification
  • Data augmentation
Explore the research

01 / MODEL

12-lead classification

Custom ECG model built around all twelve leads.

02 / DATA

Four datasets

A unified pipeline combined four varied ECG datasets.

03 / METHOD

Stress-tested signals

Noise, baseline-wander, and time-shift augmentation; ablations explored zero-shot cross-dataset generalization.

Research result, not clinical validation.0.90 peak weighted F1 · Independent research, June 2025–April 2026

02 / ABOUT

An engineer. And an explainer.

PROFILE / ANN ARBOR, MI

Building systems.
Sharing what I learn.

My work spans building learning experiences, investigating machine learning, and teaching students their first lines of code.

Download my résumé

CURRENT ORBIT / EDUCATION

University of Michigan

B.S. in Computer Science
August 2026–May 2028 (expected)

Previously: Troy High School, 2022–2026. Dual enrollment at Lawrence Technological University: Calculus 3 and Linear Algebra.

FLIGHT LOG / EXPERIENCE

01 / MAY–AUGUST 2026

Coding Instructor

iCode Troy

Designed lessons in Scratch, HTML/CSS, and Python for students aged 5–10, and introduced foundational generative AI literacy.

Making technical ideas approachable.

02 / JUNE–AUGUST 2025

Software Engineering Intern

Paraxis LLC

Overhauled an ed-tech platform’s UI and built a Brilliant-style learning flow, alongside interactive Next.js lesson modules, dynamic content, and real-time progress tracking.

Configured Supabase authentication and row-level security policies. The platform overhaul earned $100,000 in startup investor interest.

Product experience, backed by engineering.

ONBOARD TOOLKIT

  • Python
  • Java
  • TypeScript
  • React
  • Next.js
  • Node.js
  • Tailwind CSS
  • PyTorch
  • TensorFlow
  • Git
  • Playwright
  • Vitest

03 / CONTACT

Let’s make something useful.

If my work fits your team, I’d love to connect.