Skip to main content

Status: live. Delhi --:-- IST · agents running · now buildingLumora

Subhav Mathur

Software engineer building AI products and SaaS.

4+ years on healthcare backends at a large US health-tech company — now building voice agents, agent teams and SaaS products under Xcution Labs.

Shipped with tests, guardrails and a human approving what goes live.

Selected work

All products below are mine and early-stage. No fake client logos—just live code, architecture and run logs.

  • Lumora

    MVP

    Small businesses in India lose leads to missed calls and slow follow-ups. Owners don't have time to learn another tool.

    • Cloudflare Workers
    • D1
    • Clerk
    • Zod
    • TypeScript
    Read case study
  • Teamora

    Building

    Coding agents are fast but they hang, loop, take credit for each other's work and leave half-finished runs writing to the repo.

    • TypeScript
    • Node.js
    • PostgreSQL 16
    • GitHub Actions
    Read case study
  • NamasteParent

    Live, in development

    Families living away from aging parents worry every day, and parents don't want to learn a new app—they just want to answer a phone call.

    • Gemini 3.1 Flash Live
    • Piopiy/TeleCMI
    • FastAPI
    • Next.js
    • Supabase
    • Cloud Run
    Read case study
  • CivicFlow

    Open source

    Public benefit portals bury applicants in complex forms, while brittle screen-scraping bots break or submit wrong data without human approval.

    • TypeScript
    • WebMCP
    • Zod
    • MIT
    Read case study
All work, including internal builds

Lab

Small experiments and tools — including the interactive voice demo that used to sit on this page.

  • Voice agent / Interactive

    Smile Studio voice-agent demo

    The scripted bilingual receptionist console from the Lumora build — talk, type or play the 28-second sample call and watch the intent/policy/tool/receipt telemetry fill in. Honest demo: no real bookings or data stored.

  • Grok Bot / Multi-Agent

    Fleet Interviewer

    A Grok Bot that interviews you about your stage and spins up a dedicated bot team (Chief of Staff → Engineering Manager) where each bot has one job.

All experiments

Writing

A dated build log. The models are fine — the plumbing around them is where the time goes.

  1. I almost removed my plan reviewer

    How a pre-code review agent caught 6 serious issues, including a bug that would have silently wiped a to-do list of owed fixes.

  2. What went wrong with my agents: 6 lessons from their logs

    45-minute hangs, 13-round review loops capped down to 4, leftover background processes, and why the plumbing around the models is where the time goes.

  3. The voice agent is the commodity

    Why the real product isn't the voice stream—it's capturing intent, taking action, following up, and updating the business.

  4. Intent, policy, tools, receipts

    A four-layer mental model for shipping agents that never put production one bad tool call away.

All writing

About

Healthcare backends by day, voice agents and agent teams by night.

4+ years building Java/Spring Boot healthcare backends at a large US health-tech company: access controls, PHI/PII-safe logging, eligibility rules, CI/CD. Now I design the architecture for AI products, run teams of coding agents and review every change. Agents can open PRs; only I merge.

Full timeline & values

What builders said

  • [ADD TESTIMONIAL WITH PERMISSION — Voice Agent Pilot Client]
  • [ADD TESTIMONIAL WITH PERMISSION — Agent / Automation Client]
  • [ADD TESTIMONIAL WITH PERMISSION — SaaS / MVP Client]

Currently running early pilots. Want to be a launch case study?

Work with me

Alongside my own products I take on a small number of fixed-scope builds — voice agents, AI automations and SaaS MVPs. Written scope in 48 hours, weekly staging demos.

Get in touch