Prakhar Varshney — AI systems engineer

I build systems that think together.

Current coordinates
28.6139° N / 77.2090° E
Local time 00:00:00 IST

From interfaces to distributed systems to multi-agent orchestration—I turn ambiguous, human workflows into reliable software that moves at machine speed.

Scroll to trace the route ↓

The route

Three coordinates.
One evolving point of view.

Places changed the scale of the problems I could see. Drag the globe or choose a stop to follow the shift—from learning how software works to designing systems that coordinate intelligence.

Natural Earth / orthographic view
28.61° N · 77.20° E
Drag to rotate · Click a signal
Origin — 202028.61° N / 77.20° E

Delhi

Where curiosity became craft.

Delhi is both the origin and the foundation: taking things apart, asking better questions, then turning that curiosity into a computer-science education and a bias toward shipping.

  • Curiosity
  • B.Tech Computer Science
  • First products
01 / 03

How I think now

Not one clever model.
A well-run team.

The hard part of agentic software is not the demo. It is designing roles, memory, guardrails, evaluation, and fallbacks so the whole system remains useful under pressure. Explore the operating system below.

ORCHESTRATOR/ ROUTER INTAKE SPEC RESEARCH MATCHER EVALUATOR
1,200+Sourcing requests / month
95%+Spec generation accuracy
100+Specifications / minute
70%Faster turnaround

Version history

The stack grew.
So did the questions.

My career is less a ladder than a widening lens: from “can I build this?” to “should this exist, can it scale, and can people trust it?”

01

Make it real.

2016 — 2021I learned by shipping: interfaces, APIs, payments, queues, deployments. The feedback loop was immediate—users either got value or they did not.

Operating range

Wide stack.
Sharp edge.

I work across the stack because agentic products rarely fail in only one layer. My sharpest edge is connecting model behavior, distributed infrastructure, product experience, and measurable outcomes.

01

Agentic AI

Multi-agent architecture · Model routing · RAG · Evaluation & guardrails · Prompt design · Token optimization

02

Distributed systems

Serverless workloads · Event-driven architecture · Queues · Microservices · Realtime systems · Failure recovery

03

Product engineering

React · Next.js · TypeScript · Node / Nest · Python / Django / Flask · Flutter · Zod · Tailwind

04

Cloud & data

AWS CDK · Lambda · Step Functions · SQS / SNS · PostgreSQL · MongoDB · DynamoDB · Docker · Kubernetes

05

Core languages

Python · JavaScript / TypeScript · Java · C / C++ · Dart · SQL · Bash · HTML / CSS

Selected work

Proof, not
posture.

Three chapters where the system got faster, the workflow got lighter, or a previously manual process learned how to run itself. Click a chapter to inspect the evidence.

01

Sourcy Global

Senior Software Engineer · Singapore

2024 — Now
+

Built the agent harnessEngineered the production runtime behind an autonomous sourcing team: specialized agents, Zod-validated tool contracts, shared workflow state, checkpointed execution, retries, fallbacks, and human escalation across Gemini, OpenAI, and Anthropic.

Made quality measurableBuilt reliability loops around every stage with constraint and schema validation, evidence-grounded outputs, hallucination checks, confidence gates, regression evals, and model-based scoring—sustaining 95%+ spec accuracy across 1,200+ monthly requests.

Optimized every tokenReduced AI cost and latency through task-aware model routing, context compaction, token budgets, caching, batching, parallel fan-out, and early exits—reaching 100+ specs per minute and 70% faster turnaround.

Productionized the intelligenceRan the system on AWS Lambda, Step Functions, SQS, CDK, Python, TypeScript, and PostgreSQL with per-run traces, cost/latency telemetry, replayable jobs, and recovery paths powering AutoQuotation, supplier ranking, and product mapping.

Experiments that shipped

Side quests with
real users.

Prakhar Varshney reading a book in a studio, wearing a Sourcy polo shirt Prakhar Varshney / 01

The human in the loop

I became more than my stack.

Crossing cities, domains, and systems taught me that engineering is a social act. The best architecture begins with listening; the best automation preserves human judgment; and the best teams make room for dissent, ownership, and a little wonder.

Clarity over cleverness
Evidence over theatre
Humans own the outcome
Stay curious, ship responsibly

Next coordinate

Have an impossible
workflow? Let’s talk.

Prakhar Varshney
AI systems · Product engineering · Agent orchestration