SAURAV / 2026BUILDING THE SIGNAL
Observe.

AI ENGINEER · PRODUCT BUILDER · KOLKATA

I make AI
behave.

Saurav Mukherjee builds intelligent products that can explain themselves, survive production, and still feel human.

Abstract glass sculpture representing an AI system
currentlyAI Engineer @ EYenterprise agents / Azure
Kolkata, India
community built10K+ developers
drag the artifacts
A CAREER SHAPED AROUND
EY
MercorMercor
GoogleGoogle
CodeIN CommunityCodeIN
ConnectBudConnectBud
01 / FIELD NOTES

Signals from where
AI takes shape.

Six high-resolution films and one still from summits, installations, and the conversations around them. Media loads only as it enters view.

From screen to room
01From screen to roomA product story should survive outside the laptop.
02Walking the ecosystemA moving field note from the summit.
03Portraits of progressA playful interface for a serious ambition.
04Supporting India’s AI ambitionStorytelling at the scale of a room.
05The system answers backA focused excerpt from an interactive AI installation.
06Across ecosystemsA conversation from the India AI Impact Summit floor.
07Inside the circleAn interface discovered through motion.
02 / SYSTEMS

Ideas engineered into
living systems.

Three products built around the difficult part: evidence, reliability, and what happens after the demo.

01 / Governed agents

OpsWeave

Agents can improvise. Operations cannot.

A production-minded workspace that turns messy multimodal evidence into versioned, inspectable workflows with evaluation and human approval built in.

Next.jsFastAPIAWSTerraform
opsweave.sauravmukherjee.in/
OpsWeave product interface
Typed workflow contractsEvaluation gatesTraceable execution
01
02 / Environmental intelligence

VerdaTrace

A source of truth should show its working.

An evidence-first workspace that converts approvals, amendments, maps, and field records into current obligations and inspection-ready action.

TypeScriptMapLibreDocument AINext.js
verdatrace.com/
VerdaTrace product interface
Clause-level citationsAmendment graphSpatial validation
02
03 / Safety-first research

HealthAI Agent Lab

Different risks deserve different model boundaries.

An open-weight health-screening research system combining specialist ML, local voice inference, typed tools, and deterministic safety precedence.

PythonFastAPIOpen modelsAWS SAM
healthai.sauravmukherjee.in/
HealthAI Agent Lab product interface
Open-weight boundaryEmergency precedenceVersioned models
03
03 / VISUAL ARCHIVE memory index online

Proof that life
exists off-screen.

27captured frames

Every frame in the source archive, re-cut as a lightweight visual journal. Tap a frame to inspect it.

27 FRAMES / ONE CONTINUOUS FIELD NOTEtap to inspect
04 / EXPERIENCE

A career built
through the work.

Five chapters across enterprise AI, cloud-agent evaluation, cybersecurity, open source, and developer education—each organized around the work itself.

CHAPTER 01 / Sep 2024 — now

AI Engineer

EY (Ernst & Young)

I turn ambiguous enterprise requirements into governed AI systems that are secure enough to operate, observable enough to trust, and practical enough to ship.

01Architected a role-aware multimodal knowledge retrieval agent on Azure AI Foundry, Azure OpenAI, and Azure AI Search.
02Built hybrid retrieval with Dataverse and Cosmos DB, adding access controls, provenance, freshness, and audit-ready observability.
03Delivered multi-agent test generation plus governance and FinOps signals across FastAPI, Azure Functions, Power Platform, and .NET.
Azure AI FoundrySemantic KernelMCPFastAPICosmos DB
PRODgoverned AI delivery
OPERATING PRINCIPLEMake the complex legible—then make it dependable.
UNDERSTANDBUILDEVALUATEOPERATE
05 / CAPABILITY EVIDENCE

Trace the stack.
See where it shipped.

Explore the complete résumé toolkit. Hover, focus, or tap any skill to connect it to real company work and public products.

SELECT A SIGNAL
ACTIVE SKILL / VERIFIED CONTEXT
RAG

Agentic RAG

Grounded enterprise answers and compiled multimodal evidence into reviewable workflows.

USED INEY · OpsWeave
capability.trace
$ portfolio inspect --skill="agentic-rag" evidence: EY · OpsWeave context linked · proof resolved
problemagentic ragworking system
01 / REASON

Agentic AI & LLM systems

11 skills
02 / LEARN

AI / ML

12 skills
03 / OPERATE

Platform, cloud & backend

12 skills
04 / SHIP

Frontend, data & tools

9 skills
05 / EXPRESS

Languages

6 skills
PROOF OF WORK / PUBLIC GITcomplete activity