Platform Product · Enterprise → AI

I run Commercial Data & Analytics by day, and ship AI products by night.

Twenty years of enterprise-grade constraints. Three live products' worth of consumer-grade polish. The combination is the point — most people have one or the other.

Occasionally open to exceptional AI / Platform / Data Product Owner conversations — not actively looking.

Head down on Quantlys. Always happy to talk about platform, productivity, or developer-facing products.

Quantlys — multi-agent
orchestration, in practice
What I've Shipped

Three products. One AI platform.
Nights and weekends.

Each one started from a real problem somebody described out loud, not a gap on a slide. All three are live right now — you can download one, clone another, and call the third. An AI workforce either ships working software or it doesn't; this is the part that isn't a claim.

GreenBook landing page — the lawn-care job book that works with no signal
GreenBook
Landscaper CRM · Android
Live

Bilingual, local-first job book for solo landscapers. English and Spanish at one tap, and it opens and saves with no reception — because the job site has none.

Free on Google Play · No signup
Quantlys Meeting landing page — leave the call with a spec, not notes
Quantlys Meeting
Meeting → PRD
Open source

Video in a browser tab. Guests join from a link, not an account — and the working session writes a PRD instead of a recap you have to rewrite. MIT licensed: clone it, bring your own keys.

MIT · v1.0.0-oss · LiveKit · Deepgram
Tablerene landing page — AI phone host for restaurants
Tablerene
Voice agents · Restaurants
Live

An AI phone host that answers on the first ring while the team runs service. Takes to-go orders, books tables, and hands off to a human for allergy questions instead of guessing.

Voice agents for service businesses
1 · Draft in parallel 2 · Peer-rank, blind 3 · Arbiter synthesis 4 · Build & verify MODEL A MODEL B MODEL C MODEL D MODEL E RANK A–E SELF-VOTES STRUCK ARBITER WINNER-WEIGHTED RUNS THE APP 117 PERMANENT CHECKS WEB · iOS · ANDROID
Ask Me

The questions an interviewer
actually asks

Pick one. Written answers are available for every question.

The Thesis

Consumer-grade polish,
enterprise-grade constraints

Most product people have lived on one side of this line. Either they have shipped things people choose to use, or they have shipped things inside an organisation where governance, audit and procurement decide what is possible. The two disciplines teach opposite reflexes, and the gap between them is where most enterprise AI quietly dies.

I work both sides every week. By day the constraint is real: pharma commercial data, 110+ source systems, governance that has to hold. By night the constraint is equally real and completely different — if a stranger has to create an account before they get any value, there is no product.

What that combination produces is a specific instinct: ship the thing that survives a security review and that someone picks up without being told how. I have had to learn both the hard way, on the same calendar.

Skill Map

Eight skills,
rated honestly

Four I'd stake a product on, two I'm solid at, and two I'm still building in the open. Each one links to the work that backs it up.

AI product ownership Strong

Taking an AI product from a problem someone described out loud to something people use.

LLM orchestration & cost economics Strong

Choosing the right model for each job, and holding every build to a budget.

Multi-agent engineering leadership Strong

A team of AI agents and a few humans, with clear lanes and a short list of what only I approve.

Release, security & risk discipline Strong

Shipping fast without shipping something I can't take back, and saying what still broke.

Realtime voice & integrations Solid

Phone and video, live: the plumbing behind an AI phone host and a meeting that writes its own PRD.

Unit economics & pricing Solid

Let the cost driver set the price. A flat price on a usage-priced product moves the risk onto you.

GTM, SEO & positioning Building in the open

Now: learning what people actually search for and how to say what Quantlys is in one sentence. This is the skill I'm least likely to hand off.

Getting & keeping users Building in the open

Now: counting the steps between a stranger and their first win, and deleting the ones that don't earn their place.

Platform at Scale

Five product hats,
one platform

Product ownership at Gilead isn't one title for me — it's five concurrent ones, across launches, field tooling, commercial AI and site technology. This is the enterprise half of the thesis, and it is where the constraints live.

Product OwnerAI Products
Commercial AI product ownership spanning primary market research and competitive intelligence — taking generative capability from use case to something the business can actually run.
Product OwnerData & Analytics,
Product Launches
The data, targeting, alignment and readiness a pharmaceutical launch runs on. Launches do not move for the platform; the platform moves for the launch.
Product OwnerField Reporting App
IT Product Owner for the iOS app serving the entire Gilead field force — pre-call planning and call analytics in the reps' hands.
Operations LeadCommercial Analytics
Run ownership for the platform the commercial organisation reports on: release, reliability, data quality and incident response.
LeadershipSite Technology Lead
Member of the site technology leadership group for Gilead's Raleigh site.
$20M
revenue leakage stopped through automated audit reporting
7,000+
analytics users enabled across Sales, Finance, Supply Chain and HR
1,500+
workdays saved via RPA and bot implementation
Scale in the current role
$5M
portfolio delivered on budget
99%
on-time delivery
50+
specialists led, internal and vendor
2.5 PB
commercial data under management
Platform Modernization
On-prem analytics migrated to AWS, with federated query and lakehouse pipelines underneath.
Applied GenAI
Three generative pilots taken from proof of concept through governance review.
Self-Service Enablement
450+ governed data products centralized so analysts serve themselves instead of queuing.
Launch Readiness
Cross-franchise data hubs standing up targeting, alignment and performance tracking.
How I Work

Four things I had to
learn the expensive way

DiscoveryCount the steps somebody pays before they get any value, and talk to the people who get blamed when it breaks — not only the people who asked for it.
PrioritisationA new step has to remove two elsewhere. Acceptance criteria get written before the work starts, so the argument is about the spec rather than about who remembers the meeting differently.
Launch under governanceSecurity review, data governance and audit are design inputs, not a gate at the end. The version that survives review is the one that was built expecting it.
After launchDone is what the user sees. If nothing looked at a pixel, nothing was checked.
"Define done as what the user sees."
A build reported every step green, then the emulator drew a white screen. Every check verified its own homework; not one checked a pixel. The pipeline now looks at the phone.
"A guard that never fired isn't protection. It's decoration."
Two safety controls, broken in opposite directions, both passing review. Only the firing record showed it. Test the theory a control encodes, not just the code.
"Count the steps someone pays before they get value. Then delete two."
Five barriers stood between a stranger and anything useful. I'd never counted them, because in my own testing I was already past all five. Down to two, people started using it.
"Measure blind, hold out the tests, and grade the grader."
Across 675 ranking ballots, models put their own work first. My evaluator had a bug that inverted a whole conclusion. Consensus without blinding is partly self-promotion.
Experience

Twenty years of making data
trustworthy enough to act on

Gilead Sciences
Raleigh, NC · 2020 – Present
  • Associate Director, IT Data Engineering
  • Sr. Manager, IT Data Engineering
  • Manager, Data Analytics Program (Commercial)
  • Led commercial cloud transformation, migrating on-prem analytics to AWS.
  • Directed delivery across IT, Sales Operations, Digital Marketing, Global Analytics and Data Governance.
  • Progressively promoted Manager → Sr. Manager → Associate Director.
Symantec
2018 – 2020
  • Business Intelligence Program Manager
  • Led the BI Center of Excellence; drove Power BI adoption enterprise-wide.
  • Executive Sales/Finance dashboards covering ~$4B in sales, with predictive modeling.
NetApp
Sunnyvale, CA · 2008 – 2018
  • BI Sr. Domain Program Manager
  • Tableau Architect / Program Manager / BI Analyst
  • Built the Tableau / OBIEE Center of Excellence — 100+ metrics, 20+ dashboards, 200+ reports.
  • Automated on-license audit reporting, stopping $20M in revenue leakage.
Gilead Sciences
Foster City, CA · 2016 – 2017
  • Sr. Business Systems Analyst / Technical Manager
  • Established an Enterprise Tableau CoE across 15 business units.
Themes I build in public
For hiring managers

The short version,
in five lines

  • TodayAssociate Director, IT Data Engineering at Gilead Sciences, Commercial Data & Analytics — five concurrent product-owner and lead roles across product launches, the field iOS app, commercial AI, analytics operations and Raleigh site technology.
  • Scale owned110+ source systems, 2.5 PB of commercial data, 7,000+ analytics users, 450+ governed data products, 50+ specialists led across internal and vendor teams.
  • Delivery record$5M portfolio delivered on budget at 99% on-time. 1,500+ workdays saved through RPA. $20M in revenue leakage stopped at NetApp through automated license-audit reporting.
  • Shipped soloThree live products. GreenBook — bilingual CRM for solo landscapers, works with no signal, free on Google Play. Quantlys Meeting — turns a working session into a PRD instead of a recap, MIT on GitHub. Tablerene — voice agents that answer a restaurant’s phone during service.
  • Why both halvesEnterprise governance and consumer-grade adoption on the same calendar for three years. All three products run on one patent-pending multi-agent orchestration engine. Based in Apex, North Carolina; open to IC or player-coach.
Education & Certification

Education

Master of Business Administration (MBA)Kalinga University
B.Sc., Computer ScienceAndhra University
Project Management Professional (PMP)Project Management Institute
RPA Program ManagerAutomation Anywhere
Contact

Let's talk

Most interested in product roles where enterprise constraints and consumer expectations meet — platform, productivity, or developer-facing. Also happy to compare notes with anyone building solo.

Email santosh.adari@gmail.com
Quantlysquantlys.ai
Based inApex, North Carolina