Shubham Shrivastava

Where it started

I wanted to build the games I grew up playing.

As a kid I'd come home from school, drop my bag, and go straight to the screen, still in my uniform. Project IGI, GTA Vice City, Total Overdose, Hitman, Contra, Mortal Kombat. I finished all of them.

Later in college it was Rainbow Six Siege. Six to eight hours straight with friends, no concept of time.

At some point the question shifted from 'what do I want to play next?' to 'how does any of this actually work?' That took me to ICAT Bengaluru for a BSc in Game Development. Three years of Unity, C#, 3D modelling, physics systems, collision engines, shaders, animation rigs.

What the degree actually taught me: the games I'd been playing my whole life weren't made by one person with a good idea. They were the output of game designers, developers, QA teams, producers and artists, all moving together.

Building something real is a coordination problem as much as a technical one. That stuck.

Spatial computing before it had a name

A Mini Cooper in a headset, and I knew this was the next thing.

When I was interviewing at AutoVRse, my manager Akash handed me an HTC Vive and told me to try something they'd built. A Mini Cooper experience. I put on the headset. That was it. I walked out knowing spatial computing was where I needed to be.

At AutoVRse I shipped as an individual contributor. An AR brand game and an enterprise VR experience.

The game ran for a global CPG brand's 75th anniversary. QR codes hidden around the office, one letter each. Scan your assigned product and a 3D alien spawns. You throw chocolates at it. The chocolates it eats are your score.

Seven days, daily winners, one winner at the end. It shipped to Google Play and the App Store. It taught me how different shipping for a headset is from shipping anything else. Hardware dependency makes every assumption expensive.

Then I got to know about Microsoft HoloLens and Mixed Reality. I tried it. That was the Tony Stark moment. Holograms in real space, persistent in the physical world, interactive without a controller.

If the Vive told me spatial was next, HoloLens told me where it was actually going. That's why AfterNow happened.

AN-Prez, and what we decided it shouldn't be.

It wasn't slides in a headset. It treated the room itself as the canvas. You broke a presentation into scenes and placed 3D models, images and video freely in physical space.

Location triggers let you anchor content to a real object, so walking up to a physical product started the section about that product. The environment became the stage instead of the backdrop.

Transitions could run fully automatic for an unattended showroom, or entirely on the presenter's command when the room mattered. And it was multi-user: presenter and audience in different physical locations, sharing the same holograms.

It wasn't a demo. AN-Prez shipped publicly on the Microsoft Store and ran live with Fortune 500 clients across healthcare, aeronautics and telecom, who used it to distribute and manage presentations across their own device fleets.

It later extended beyond HoloLens to Oculus Quest, iOS and Android. AfterNow was an official Microsoft Mixed Reality partner, and the work was covered by VentureBeat, TechCrunch, Mashable and Variety.

The moment the product got good was the moment we stopped asking how to show a slide in 3D, and started asking what a physical room can do that a screen can't.

We were building spatial computing before anyone called it a product category. When Apple announced the Vision Pro in 2023, it was the first time I felt like I hadn't been pointing in the wrong direction.

The pivot that changed everything

I discovered I was more interested in the problem than the code.

At AfterNow I worked alongside Simi Shenoy, who was gathering requirements from clients before handing them over for execution. Sometimes we'd sit together and dig into what the client actually wanted to solve for their end users. That was my first real exposure to product thinking.

I noticed something. When requirements were frozen and it was time to build, I was fine. But the conversations before that, understanding the problem and untangling what the client said they wanted from what they actually needed, intrigued me far more than the coding itself.

That realisation led me to Avataar, where I joined as an Associate Product Manager. Three years writing code became five and a half building product. Somewhere in the middle of it I took an MDP in Data-Driven Product Management at IIM Lucknow, 2021 to 22.

Five years at Avataar. Three titles, four products I owned end to end.

The human editing in Creator produced the template corpus Velocity's matching algorithms later ran on. That's what makes these four products one campaign rather than four entries.

Incarnate is the one that taught me the most, because of where it broke. Simple objects came out clean, and a few minutes of cleanup made them production-ready. Botanicals and food failed badly.

Anything very small was hard to scan, anything very large was harder, and complex bending was close to impossible. It worked beautifully inside certain categories and certain sizes, and not at all outside them.

Learning to say that out loud, this works here and it doesn't work there, was more useful to me than any launch.

All four sat inside the same core problem: how do you make physical products come alive digitally?

Then the work that defined the second half.

The Agentic AI Platform, taken 0 to 1 and validated through enterprise POCs in medical coding. The brief was unlike anything I'd built before. Take workflow requirements in plain English, deploy AI agents to create the workflow, and achieve the desired goal autonomously.

Not a chatbot. A system that reads intent and builds the automation itself. Multi-agent orchestration before the term was mainstream.

A new problem worth five more years

Five years on product visualisation. Time to find the next hard thing.

After five years solving the same category of problem, how to show a physical product in the most compelling way possible, I wanted a domain where the complexity was different, the stakes were higher, and the workflows had been broken for longer.

Healthcare or finance. Both excited me. Finance won.

Neoflo came at exactly the right moment. A $10M seed round from Lightspeed and Peak XV. A small founding team with a clear thesis: back-office operations had not been meaningfully touched since ERPs were invented.

Accounts payable, accounts receivable, reconciliation, expense approvals. Decades of manual work, duct-taped together with spreadsheets and workarounds that everyone had just learned to live with.

My first week I sat with an accounts payable manager and watched her work. 47 browser tabs. 3 spreadsheets. An ERP that looked like it was built when BlackBerry was cool. And a folder on her desktop called INVOICES_FINAL_FINAL_v3.

She wasn't bad at her job. She was good at it. She had built an elaborate system of workarounds to compensate for software that had never been designed around how she actually worked.

That's the problem. Not 'finance teams need AI'. The tools never matched the workflow, and everyone quietly absorbed the difference.

I joined as Head of Product and a founding team member. Another 0 to 1. The difference this time: the problem isn't about making things look better. It's about making entire workflows disappear.

AI that actually does the work. Invoices matched, exceptions flagged, payments routed, without a human in the middle of every step. Live with enterprise clients now, one of them running 6,000+ invoices a month across six countries. The chaos is familiar.

Every move started with a feeling, not a plan.

I didn't set out to work on cutting-edge technology. I put on a headset during a job interview and walked out a convert. I stumbled into product management while watching a colleague ask better questions than I was asking.

I left Avataar not because something better appeared, but because five years on one category of problem felt like enough, and I wanted to find out what a genuinely different kind of hard felt like.

The throughline isn't the technology. It's the pull toward problems that haven't been solved yet.

Spatial computing when nobody had a consumer headset. Agentic AI before the term existed. Back-office automation that's been manually duct-taped together since ERPs were invented.

I came up as a developer, which means I still think in systems. I write code. I sit with the workflow before I spec the solution. That's not a habit I'm trying to change. It's the thing that makes me useful at the beginning of a problem, when the shape of it isn't clear yet.