Product & Venture Leader · Enterprise AI · Complex B2B

I translate fragmented, expert-heavy problems into products that ship, scale, and compound.

Over six and a half years at UPMC Enterprises, I turned unfamiliar clinical workflows, from cognitive care to chronic pain, into products, ventures, and investment calls that ship, scale, and compound. I run the same playbook today building AI-native products myself, agentic systems included.

SEE THE WORK → GET IN TOUCH
Debashish Sasmal
THE PATH ·

Eight and a half years in engineering roles before venture leadership at UPMC Enterprises, plus guest lecturing at Carnegie Mellon's MS Product Management program and co-hosting The Product Signal.

// THE PROOF

Three ventures, the same method.

A venture built from zero, an EHR-scale investment, regulated healthcare. Different bets, same pattern: find the real constraint, then re-architect around it.

01 · 0→1 VENTURE · CHRONIC PAIN
0→1
concept to build
100+ PCPs
single-signature access
Chronic-Pain Care Venture
A new care-model thesis, built from first principles into an enterprise adoption path.
View case study → 02 · INVESTMENT · EHR INTEGRATION
$3-6M (est.)
recoverable reimbursement/yr
90% ↓
onboarding time
Cognitive Screening Platform
The constraint was workflow, not science. Integrated into the market-leading EHR.
View case study → 03 · PLATFORM → STRATEGIC ACQUISITION
≈92%
response time ↓
$6–9M/yr
est. system savings
Acute-Care Virtual Platform
An internal workflow tool repositioned as a sellable, multi-tenant asset.
View case study →

// HOW I LEARN A DOMAIN

I don't start with the product. I start with the room.

Rapid domain acquisition is the skill that lets me move between markets. Here's the sequence I run every time.

01
Map the experts and their incentives

I learn who holds the knowledge, who decides, who pays, and who quietly blocks. In expert-heavy markets, the org chart lies and the incentive map doesn't.

02
Hunt the constraint, not the feature

Almost every product I've shipped won because I found the real bottleneck (specialist time, onboarding friction, a broken handoff) and built against that.

03
Write a thesis I can be wrong about

I put a falsifiable thesis on the table early and let domain experts stress-test it. The fastest way to learn a market is to be wrong in front of people who know better, then fix it.

04
Decide build / buy / partner from the thesis

Once the constraint is clear, the path usually is too. My job is to make that call honestly, even when it's inconvenient for the org's default.

I'm running this same sequence right now on Keystone AI, an agentic build system gated by exactly this kind of judgment call at every stage.

See the AI builds →

// THE THROUGH-LINE

The constant was never a domain. It was a method.

Same method, three markets: CRM data, real estate, and six years in regulated healthcare, where the stakes and the expertise run deepest.

More about me →

// EXPLORE

// WORK

Case Studies

Three ventures across digital health: the bet, the call, and what it proved.

EXPLORE →

// AI

AI Builds & POV

A point of view on AI in regulated products, backed by things I actually built.

EXPLORE →

// ABOUT

Who I Am

The through-line, the method, the podcast, the teaching, the track record.

EXPLORE →

// CONTACT

Let's Talk

Building something ambitious in a hard market? That's the conversation I want.

GET IN TOUCH →