AI Infrastructure Engineer · Agent Systems Builder · Founder at SuperMind

I build the systems that let agents work.

About

I’m a 4th-year BTech student and founder who builds at the intersection of AI systems and real products. I started Lucentra Labs with and shipped SuperMind, GoverRAJ, and PenguinBot. My work is about one thing: making agents actually work in the real world, not just in demos.

945

GitHub commits · last year

#4

Product Hunt peak

“I don’t just build AI apps. I build the systems that let agents work: context, tools, execution, control, memory, and debugging.”

Featured Work

Selected projects

01

SuperMind

Founder · Architecture & Infra

Run your service business, SaaS product, commerce brand, or internal ops company from one SuperMind deployment. Each business gets isolated knowledge, connectors, approvals, budgets, and audit trails, while you keep one founder-level control plane across the whole portfolio. It works like a team of 13 AI agents automating the company with human approval wherever it matters.

50+

Active Users

13

AI Agents

1

Control Plane

Agent RuntimeContext EngineTypeScriptNext.js
02

Agent Prism

Builder · Tracing

Drop-in tracing for agent pipelines. Local SQLite, real costs, and a dashboard bundled with the npm package.

Local

SQLite

Real

Costs

Bundled

Dashboard

TracingSQLiteDashboardTypeScript
04

GoverRAJ

Lead Engineer

Police promotions in Rajasthan used to take 6 to 14 months — everything handwritten. Roster files, seniority lists, exam records, 8 manual steps, all prone to errors. I built GoverRAJ to replace that entire pipeline. Input the data, apply 30+ government rules automatically, output everything. What took months now takes weeks.

6–14mo

Before

2–4wk

After

8

Steps Automated

Rules EngineRAGPythonAES-256.NET 8
05

Aura OS

Creator · Systems

An ambient agent desktop where memory, tools and execution live on one surface. Instead of bouncing between apps, agents operate alongside you — holding context, running tasks, and surfacing the right tool at the right moment.

1

Surface

Local

First

ElectronLocal AgentsMemoryIPC
06

AgentTraceCorder

Builder · Debugging

Debugging agents is impossible when runs aren't reproducible. AgentTraceCorder records, replays and debugs agent runs deterministically — so you fix what actually broke instead of chasing flaky behavior.

Deterministic

Replay

0

Flaky Runs

ReplayDeterminismTypeScript

Also Built

Side experiments

Smaller builds, research spikes and hackathon projects — where I test ideas before they grow into something bigger.

  • 01Garuda Lens
  • 02WildEye
  • 03Hackathon Projects

Skills

What I work with

The craft stack I reach for.

Agent Systems

Planning, tool use, memory and multi-step control that make agents reliable beyond demos.

  • Context Engineering
  • Tool Use
  • Execution & Control
  • Memory
  • Multi-agent Orchestration

Harness Engineering

Tracing, evals and replay systems that expose where complex runs break and why.

  • Tracing & Observability
  • Evals
  • Deterministic Replay
  • Debugging

Product Engineering

Interfaces, APIs and product surfaces that turn system capability into something people use.

  • Next.js
  • React
  • TypeScript
  • 0→1 Product
  • API Design

Infrastructure

Queues, data, retrieval and deployment plumbing that keep production systems moving.

  • Pipelines
  • RAG
  • Postgres
  • Queues
  • Vercel / Cloud

Desktop / Enterprise

Local-first workflows, desktop shells and enterprise-grade orchestration surfaces.

  • Electron
  • Local-first
  • IPC
  • Enterprise Workflows

Open To

  • AI infrastructure teams
  • Agent systems work
  • Full-stack 0→1 product engineering
  • Founder / operator collaborations
  • Remote · Hybrid · Relocation