Currently AI Agent Developer at TestAutomate

Sanjay Kumar S

AI Agent Engineer

I build AI agents that take real actions in real systems: browser-testing agents, voice agents that book appointments, and streaming detection systems. Final-year CSE at VIT Chennai.

Chennai, India

Shortlisted, Entrepreneurs First Bangalore (Fall 2026)
EFShortlisted, Entrepreneurs First Bangalore (Fall 2026)
GitSlice, open-source Go CLI
21★GitSlice, open-source Go CLI
Commits on a solo SaaS, shipped live
297Commits on a solo SaaS, shipped live

01 / Experience

Where I work

  1. AI Agent Developer · TestAutomate

    Jul 2026 – Present

    Part-time · Remote

    • AI Agent Developer at TestAutomate (testautomate.ai), an AI-powered QA platform that helps teams automate software testing. I work on the product's AI features, backend and web app, from building features to debugging and deployment.
  2. Software Engineering Intern · Hard n Soft Technologies

    May 2026 – Jul 2026

    Internship

    • Built browser automation workflows, Chrome extension features, AI-agent integrations and voice AI features for an AI-powered software testing product.

02 / Built for startups

Systems I built for other companies

Voice AI I designed and shipped for early-stage startups, separate from my personal projects.

Pebble

UK fintech startup

When
Mar – Apr 2026
Time spent
Several weeks

Voice intake agent for business funding

A voice agent that interviews a business about its funding needs, extracts structured details like company and revenue from the conversation, and generates the matching document checklist.

  • Structured data extraction
  • Auto-generated checklists
  • Live demo
  • Whisper
  • Groq
  • Cartesia
  • FastAPI
  • Next.js

Gensail

US startup

When
May 2026
Time spent
5 days

Real-time voice agent

A real-time voice agent deployed on the company's own server so their team could use it. Streams audio over WebSockets through a full speech pipeline, with voice activity detection for natural turn-taking and interruptions.

  • Deployed on the company's server
  • Real-time STT → LLM → TTS
  • Interruption handling
  • Pipecat
  • Silero VAD
  • NVIDIA Nemotron STT
  • Groq Llama 3.1
  • Kokoro TTS
  • FastAPI
  • Next.js

03 / Projects

Things I've built

My own products, research and open source, with the numbers from their repos.

Capstone

MediConnect AI

Aug – Oct 2026

A voice agent that finds doctors, reads out real free times and books appointments, by browser microphone or by phone call. Voice and web bookings run through the same service layer, and a doctor's live status change updates every open dashboard.

  • 7 agent tools
  • 30 concurrent bookings → exactly 1 succeeds
  • 67 tests
  • Phone calls via Twilio
  • Pipecat
  • Groq
  • NVIDIA ASR
  • Kokoro TTS
  • FastAPI
  • Supabase
  • Next.js

Research · team of 2

FraudSim

Aug 2026 – Present

Streaming fraud-ring detection on transaction graphs. An incremental engine re-checks only the part of the graph each micro-batch touches, and hub-aware weighting recovers rings hiding behind busy shared devices. Live dashboard explains every alert.

  • 36 ms per batch at 100k users
  • vs 22.5 s full recompute
  • Ring F1 0.73 vs 0.21 baseline
  • Synthetic data, p < 0.001
  • Python
  • Neo4j
  • Redis Streams
  • React
  • Docker

Solo SaaS

Skyler Inbox

Jul 2025 – Oct 2026

A Gmail organizer with custom filtered inboxes, on-device email classification and one-click summaries and replies. Ran live for several months until Google's CASA audit rule capped it at 100 users.

  • MVP in 42 days
  • 297 commits, solo
  • On-device models
  • Next.js
  • Express
  • Prisma
  • Transformers.js
  • Groq
  • LangChain
  • Supabase

Open source

GitSlice CLI

Jul 2025

A Go command-line tool that downloads a single folder or file from any GitHub repository using git sparse-checkout, instead of cloning the whole repo.

  • 21 GitHub stars
  • go install

Take-home builds and smaller projects

Short assignments for startup hiring processes, plus side projects.

04 / Skills

What I work with

Languages
PythonTypeScriptJavaScript (Node.js)GoJavaC++
AI & Agents
LLM orchestrationTool callingMCPMulti-provider LLMsTransformers.jsLangChain
Voice AI
PipecatLiveKitWebRTCWebSocketsTwilioSTT / TTS / VAD
Backend & Data
FastAPIExpressPostgreSQLSupabaseNeo4jRedis Streams
Frontend & Infra
ReactNext.jsPlaywrightChrome extensionsDockerGitHub ActionsAWSVercel

05 / Profiles

Find me elsewhere

06 / Contact

Building something with AI agents or voice? Let's talk.

Email is the fastest way to reach me, or book a call below.