Open to remote and hybrid roles

KeshavKunver

Los Angeles

AI Engineer building reliable AI-native products from 0 to 1.

I build production LLM apps: agents, RAG, and the evals that keep them from breaking.

Get in Touch View Work Download resume (PDF)

I help non-technical people ship the thing they've been putting off. Build with me

Off the clock▾

My NBA dreams didn't pan out, so I redirected the obsession into software.

NBA Finals Prediction: 76ers vs Lakers · 7/26/26

Bron, Jaylen Brown, and Embiid (when healthy) sound unstoppable
Luka has time to reset and more room to play like himself now that LeBron's gone
Philly's been starving for a Finals run, and the Lakers can be the dark horse even with the Thunder and Spurs heavily favored
#TrustTheProcess has finally come

LinkedIn GitHub Email

200+

Users

Eternal Dharma, at peak

Products Shipped

1M+

Views

Short-form social content

Years Experience

The Work

A selection

Current focus

FourthSpace

Problem

Social media isn't social. People scroll for hours and end up lonelier than when they started.

Built

A social web app with an AI agent that plans real-world meetups from natural-language chat. Built on a Claude tool-calling agent with multi-turn confirmation and safety guardrails, a real-time voice AI moderator (Pegasus, on LiveKit), and ~20 generative endpoints. Backed by a unit-test suite that encodes past agent failures as regression tests.

Outcome

MVP live.

Next.jsReact 19TypeScriptSupabaseVercelAnthropic ClaudeLiveKitPython

Website→Read the write-up→

Eternal Dharma

Problem

People want spiritual guidance but don't know where to start or how to apply ancient wisdom to modern life.

Built

An iOS app giving personalized, scripture-grounded spiritual guidance. A persona-constrained LLM assistant (Claude) with a scripture-sourced system prompt and citation rules, backed by structured verse and commentary data.

Outcome

Grew to 200+ users at peak

React NativeExpoTypeScriptFirebaseClaude (via OpenRouter)

View App→

HoopIntel

Problem

Basketball stats are scattered and hard to read. Coaches, scouts, and fans want clear insights without digging through raw box scores.

Built

A web app that turns player and team stats into visual, digestible analysis for coaches, scouts, and super-fans.

Outcome

Live web app

Next.jsNode.jsD3.jsChart.jsMongoDB

Website→

Capabilities

AI Engineering

Production LLM apps: tool-calling agents, RAG, structured outputs, and real-time voice AI. Hands-on with Claude and OpenAI APIs, the Vercel AI SDK, and LiveKit.

Core focus

Reliability & Evaluation

Eval suites and regression tests that turn past agent failures into named tests, so the same bug can't ship twice. Six years of SDET discipline, pointed at non-deterministic AI.

Full-Stack Delivery

Next.js, React, React Native, TypeScript, Python. Supabase, Firebase, Vercel. From idea to a live, monetized app.

Industry

Software Development Engineer in Test (SDET)

Infosys

Mar 2021 - Present

Design and maintain UI, API, and CI/CD test-automation frameworks that improve coverage and release stability
Cut regression effort 40% and related API defects 25%
Integrating AI tooling into the testing workflow

JavaREST APIsMySQLReactTypeScriptCI/CDAI

Software Engineer

Prathibha Management

Aug 2019 - Jan 2021

Shipped full-stack Django platform connecting LA talent with Hollywood casting opportunities
Built talent-matching system that helped performers land real roles

PythonDjangoPostgreSQLFull-Stack