Work Journey Thinking Contact Download CV
Product · AdTech & AI

Deven
Shah.

Deven Shah

Leading Product at Voiro

Nine years in AdTech, last five as a PM. I drive product strategy and roadmap with the leadership team across campaign management, yield optimisation, OMS, and AI-powered revenue intelligence for publishers and retailers across India and MENA.

For the last two years my focus has been AI products, specifically applied to AdTech. At Voiro I built and shipped Genie, a conversational data analyst built on RAG, live in production with enterprise clients. I also independently ship AI tools using the Claude API, from hypothesis to GitHub to Vercel. The hard part isn't the AI. It's making it trustworthy enough that people actually rely on it.

Product ops at Media.net, where I learned how ad serving breaks before I learned how it works. Revenue desk at Disney+ Hotstar during IPL, with ten people waiting on the number you're building. Now at Voiro, setting strategy and shipping across both India and MENA.

Let's talk → Download CV LinkedIn
Background

Where it started

2017
—
2019
Media.net
Where I learned what breaks

Joined as a product ops analyst. Spent two years on ad server migrations, market launches, and SQL queries before writing a single PRD. Watching a bid pipeline fall over at 2am teaches you more about ad infrastructure than any course. I understood the plumbing before I started designing the house.

2019
—
2021
Disney+ Hotstar
Revenue data at IPL scale

Moved to the ad revenue desk covering entertainment and sports, one of the few roles spanning the full P&L. Built a real-time day-on-day revenue tracker used by senior leadership during IPL 2020 and the Cricket World Cup. Cut reporting turnaround by 70%. If your numbers were wrong, the VP of Revenue found out before you did. That tends to focus the mind.

2021
—
Now
Voiro
The full lifecycle, every sprint

Joined as PM, now leading product. 50+ features shipped. Brand Center built from a blank document to closed beta with five premium publishers. Carrefour MENA onboarded: 164 advertiser leads, 1,300+ line items, 6M AED in Q1. Small team, no buffer between decisions and consequences.

Case studies

Selected work

01
AI Product · Voiro Genie
A conversational data analyst enterprise clients actually use
The data to spot at-risk advertisers existed. It just lived in three systems that didn't talk to each other, and nobody had time to join it manually every day.
+
The actual problem
  • Spend drop, fill rate decline, booking slowdown. All in the data, invisible to sales
  • Revenue signals siloed across GAM, the OMS, and Salesforce with no join
  • AMs pulling reports for two hours a day that should have been a morning digest
  • No scoring model. Every account looked equally healthy until it wasn't
What I built and why
  • Defined roadmap for Voiro Genie, a RAG-based conversational analyst. Started with the sales use case because the pain was clearest there
  • LLM-powered weighted scoring across spend velocity, fill rate trajectory, and booking cadence. Surfaced as a ranked list, not a dashboard
  • Pushed scored signals into Salesforce so AMs saw them where they already worked
  • Prototyped the full workflow with two AMs in Figma before writing a single ticket
What changed
  • 30% improvement in pipeline conversion. At-risk accounts re-engaged before they left
  • AM prep time for weekly reviews cut from about 2 hours to under 20 minutes
  • Sales team moved from reactive to working a call list based on actual signals
  • Genie is live in production, used by enterprise clients daily
02
OMS & Reconciliation · Voiro
Rebuilding the revenue engine while clients were already live
Genie is only as good as the data underneath it. This is the work that made that data trustworthy. CPM, CPD, and CPW pricing across India and MENA with no unified reconciliation, finance finding errors before product did.
+
What was actually broken
  • Three pricing models with no reconciliation layer tying them together
  • India and MENA going live simultaneously with different market rules in the same codebase
  • Sales closing Carrefour deals faster than the platform could model them
  • Finance catching fill rate gaps and billing mismatches. Product wasn't catching them first
The call I made
  • Rebuilt the line-item architecture rather than patch it. Slower short-term, the only real option by client three
  • Pushed validation guardrails upstream so errors surfaced at booking, not invoicing
  • Shipped Criteo and DV360 integrations. 4x data coverage across channels
  • Killed one-off client customisations and enforced reusable workflow patterns
What changed
  • Carrefour and Myntra onboarded on schedule across two markets
  • Ops team handling more clients without adding headcount
  • Finance stopped finding errors product should have caught first
  • Third and fourth clients onboarded onto the same architecture, no rework
03
0 to 1 · Brand Center
A self-serve media buying platform with no template to follow
Publishers running direct ad sales over email. No existing product doing this well for the Indian publisher stack. Started with a blank doc.
+
What made it hard
  • Two-sided product. The advertiser workflow and the publisher workflow have almost nothing in common
  • Every publisher had a different ops setup. Couldn't copy one and generalise
  • First three discovery calls gave us three different problem definitions
  • No benchmark. Had to define what "good" looked like before building toward it
How I approached it
  • Discovery across six publisher teams before writing a single requirement. Patterns only emerged at interview five and six
  • Prototyped the advertiser flow and publisher dashboard separately, validated each before joining them
  • Deliberately scoped out self-serve campaign editing in v1. The instinct was to build it, the right call was not to
  • Recruited beta partners early: Zee5, JioCinema, Paytm, PhonePe, Rooter
Where it landed
  • Closed beta with five premium publishers. PMF signal before the full engineering build
  • Direct deal workflows that ran on email now running on platform
  • Scoping what not to build saved at least two sprints
  • Foundation of Voiro's direct ad revenue product line
Product thinking

How I work

01
I test in production before I spec
When testing a hypothesis, I write code and deploy on Vercel — sometimes in an afternoon. For UI-heavy ideas, Lovable gets me a working prototype faster than Figma would. Either way, I collect feedback before engineering touches it. The spec describes something already proven, not something we hope will work.
Claude Code · GitHub · Vercel · Lovable · Figma
02
The hard part of AI products isn't the AI
Getting an LLM to produce a good demo answer is easy. Getting enterprise users to trust it enough to act on it daily is the actual product problem. Most AI failures I've seen aren't model failures. They're trust failures.
03
The second client is the architecture test
Anyone can build something that works for one client. The question is whether the decisions you made in month one hold when client two has different pricing rules, a different market, and a different ops team. I've been on both sides of that at Voiro.
04
Scope cuts are product decisions, not project management
The hardest call in a 0-to-1 build isn't what to include. It's what to leave out and defend that to stakeholders who want everything. I've gotten it wrong. I've also gotten it right.
Personal

Outside work

Cricket

IPL is not just a work event at Voiro. I have opinions about the batting order that have nothing to do with fill rates. Ask me about any India Test series and clear your schedule.

Pickleball

Picked it up recently. Faster than it looks, more tactical than people give it credit for. Currently trying to make my third shot drop consistent. It is not going well.

Travel

Need to move. New cities, unfamiliar food, conversations with people who have nothing to do with advertising. Some of my better product ideas have come from airport terminals at odd hours.

Black coffee

Made the switch from cappuccino. Haven't looked back. Mumbai has a better third-wave coffee scene than most people outside Mumbai realise.

Experience

The journey

Sep 2024 — Present
Leading Product
Voiro Technology

Full roadmap ownership across OMS, CRM, programmatic integrations, and the agentic AI workstream. Carrefour MENA was the headline: 164 advertisers, 1,300+ line items, 6M AED booked in Q1. Media Planner v1 shipped. The decisions being made now will show up in the product two years from now.

Apr 2021 — Aug 2024
Product Manager
Voiro Technology

50+ features shipped. Brand Center taken from zero to closed beta with five premium publishers. Mixpanel instrumentation across five enterprise clients. First time owning a product end-to-end. Learned what that actually means.

May 2019 — Mar 2021
Revenue Products & Analytics
Disney+ Hotstar

Built the real-time revenue tracking product used by leadership during IPL 2020 and the Cricket World Cup. Cut reporting turnaround by 70%. Ran A/B tests that moved campaign performance by 11%. Having senior stakeholders rely on data you're building is a fast education.

Jul 2017 — Apr 2019
Product Operations Analyst
Media.net

Ad server migration from external to internal infrastructure. SQL, Hive, market launches, Singapore expansion. Where I learned how ad serving actually works, not conceptually, but at the level where things go wrong.

Live work

Shipped independently

● Live on Vercel
Release Notes Automation Tool

CSMs were spending 30 to 40 minutes per release manually writing client-facing notes from Jira tickets. Built a tool that does it in under 2 minutes. OTP auth, AI-generated notes pulled from Jira, inline editing, PDF and email export. Scoped, built with Claude Code, pushed to GitHub, deployed on Vercel. Idea to live in under a week.

releasenotes-one.vercel.app →
Podcasts

In Conversation

Building AI for Ad Businesses: Voiro Genie
The Voiro Podcast · S6 E11 · Oct 2024
Building AI for Ad Businesses: Voiro's Journey with Genie

With Kavita Shenoy (CEO) and Ankit Richariya (Engineering). We talk through how Genie went from an internal idea to a production tool: the RAG architecture, the move from instruction-based to agent-based AI, and why getting enterprise users to trust the output was harder than building it.
IPL, Ads, Dreams & Data
The Voiro Podcast · S7 E3 · Apr 2025
IPL, Ads, Dreams & Data: Decoding Modern Media Monetization

With Kavita Shenoy and Abash Sinha. CTV pricing models, Dream11's creative strategy, quick commerce advertising, and what the business of media actually looks like during IPL.

Let's
talk.

Open to Head of Product or Senior PM roles in AdTech, retail media, or AI-powered products. Mumbai and Bangalore. Open to hybrid.

devenshah4172@gmail.com linkedin / deven-shah-635210145 +91 72084 20746
Download CV →
Open to

Head of Product or Senior PM at a company doing something real in AdTech, retail media, or AI-powered monetisation. Interested in 0-to-1 builds and platform-layer products. Not interested in roles where product is a ticket-writing function between business and engineering.

AdTech Retail Media AI Products Agentic AI 0-to-1 Head of Product