Bringing AR commerce to scale @ Scapic and Flipkart
Before I worked deeply in AI, I spent years building another kind of interaction shift: helping people understand products through 3D, AR, and immersive systems.
At Scapic and later Flipkart, I worked on products that helped users better evaluate what they were buying - how something looked, fit, or behaved in context - and helped businesses improve engagement, increase sales, and reduce returns. What started as a spatial computing startup eventually became a commerce platform used at massive scale.
This is where I learned that new interfaces do not win because they are novel. They win when they reduce uncertainty.
TL;DR
- Joined Scapic as a founding engineer to build web-based 3D, VR, and AR products
- Helped evolve the company from a creation platform into a B2B2C commerce business
- Built enterprise projects for clients including Airbus, Royal Enfield, Celio, ICICI Bank, and Sony Liv
- Scapic was acquired by Flipkart
- Led a team of 8 engineers post-acquisition to bring 3D and AR commerce to millions of users
- Shipped 3D and AR experiences across Flipkart, including large appliances, virtual try-on, and partnership-led AR launches
- Drove 40% engagement increase, 30% sales lift, and 10% reduction in returns
- Helped reduce Aircraft on Ground time for Airbus by ~20%
Context
The original problem was simple: people struggle to buy products confidently online.
Images and spec sheets only go so far, especially for furniture, appliances, vehicles, and configurable products. The real questions are usually spatial and experiential:
- How does this look in my space?
- Does it fit?
- Is this the right variant?
- Can I trust what I'm seeing?
That problem pulled me into web-based 3D and XR early, well before the category had matured.
Scapic: building the early platform
I joined Scapic in 2018 as a founding engineer.
The company's initial thesis was that creating 3D and VR experiences should eventually become as accessible as creating images or video. We were building web-based tools that let users:
- compose 3D scenes
- generate 360 walkthroughs
- create product viewers
- export immersive assets to social and web surfaces
Technically, that meant experimenting aggressively across Three.js, Babylon.js, A-Frame, and raw WebGL. Product-wise, it meant learning where immersive creation was genuinely useful and where it was still too early.
The early challenge was adoption. People liked the idea, but many did not want to be first. That made it important to keep the product flexible while staying close to real customer problems.
Scapic Labs: enterprise as a proving ground
To learn faster, we created Scapic Labs, a smaller enterprise arm that worked with select customers and brought those learnings back into the core product.
This phase became a proving ground for where immersive systems created actual value.
Airbus
For Airbus India, we built a HoloLens + video conferencing tool for aircraft maintenance workflows.
Ground technicians could inspect dents in engine cowls, measure dent severity, determine whether the damage was within acceptable limits, and bring in remote experts with the relevant metrics overlaid in real time.
This mattered because unresolved aircraft issues often required specialists to be flown out before the aircraft could be cleared. The result was a meaningful operational win: the system reduced Aircraft on Ground (AOG) time by ~20%.
Royal Enfield
For Royal Enfield, we built a 3D configurator for the Meteor launch and later broader catalog exploration.
One of the most important lessons from that project was that realism is not only a rendering problem - it is a trust problem. Product, engineering, design, and brand all need to agree that the digital representation is credible. In practice, that meant working closely with their product studio to match finishes, paint, and material behavior well enough that the digital experience felt reliable.
The work was successful enough that Royal Enfield expanded it across more of the catalog.
Other enterprise work
Scapic Labs also included projects such as:
- virtual try-on for Celio
- immersive spaces for ICICI Bank
- AR experiences for Sony Liv
These were all signals that immersive systems were more than novelty. In the right context, they solved real business problems.
The pivot: from creation tools to commerce
The biggest strategic shift came when Apple introduced USDZ and AR Quick Look.
That changed the distribution model. Suddenly, if you could generate the right 3D asset and optimize it correctly, users could place products directly into their space from devices they already owned.
That was the moment we pushed Scapic from a broader immersive-creation platform toward a B2B2C commerce product.
We moved quickly on two fronts:
- building a pipeline to convert heavier 3D assets into lightweight web- and Apple-compatible outputs
- testing the commerce use case with early brands like GreenSoul and WakeFit
The signal was clear. If users could see products in context before buying, conversion improved and returns dropped.
That insight reshaped the company.
Scapic was eventually acquired by Flipkart, which gave us the chance to bring these systems to a much larger user base.
Flipkart: shipping to millions
After the acquisition, the work changed completely.
The problem was no longer whether 3D and AR commerce could work. The problem was how to make it work inside a large production system, across millions of users, thousands of products, and highly fragmented devices.
I led a team of 8 engineers as part of what became Flipkart Camera and later Flipkart Innovation Lab.
We worked across three major areas:
- large appliances and home products
- beauty and cosmetics virtual try-on
- brand and partnership-led AR launches
Large appliances and home products
This was the flagship rollout.
The idea was straightforward: bring 3D and AR product viewing to categories where confidence matters most, such as refrigerators, washing machines, furniture, and similar products.
The hard part was delivery.
On the engineering side, we had to maintain performance for WebGL-based 3D rendering inside a hybrid app, on mid-range and low-end Android devices.
On the operational side, the real challenge was content. To make the experience meaningful, we needed a scalable way to generate a huge catalog of 3D assets.
We approached that in two ways:
- built and managed a pipeline of roughly 200 3D designers
- developed a photogrammetry pipeline using structured 360-degree capture from product photography teams
That dual strategy allowed us to support both:
- 360-degree product viewing from images
- generation of 3D assets for web and AR experiences
By February 2021, 3D and AR were available across large appliance listings on Flipkart.
Cosmetics: virtual try-on
We partnered with ModiFace to bring virtual try-on into the Flipkart app.
The interesting technical challenge here was integration. We needed to bridge a native SDK into a React Native / hybrid environment with two-way communication and without degrading performance. That meant working carefully across platform boundaries, not just adding a feature.
Partnerships: Snap Camera Kit
We were also the first to integrate Snap Camera Kit on an Indian e-commerce platform.
This enabled AR-led launches for products like upcoming phones and shoes. In isolation, these launches were smaller than the large-appliance rollout, but strategically they showed that Flipkart could support brand-driven AR experiences as a repeatable capability.
Outcomes
The work delivered clear business results:
- 40% increase in engagement
- 30% increase in sales
- 10% reduction in returns
- ~20% reduction in Aircraft on Ground time for Airbus
- rollout of 3D and AR across large appliance listings on Flipkart
- managed a team of 8 engineers
- managed and operationalized a pipeline of roughly 200 3D designers
- presented innovations at NRF 2019 and 2020
What I'd do differently
I would have pushed harder on photogrammetry earlier.
The manual 3D pipeline gave us quality, but it could not keep up with the speed of catalog growth. We knew photogrammetry would eventually be the more scalable path. In hindsight, I would have invested more aggressively there earlier and accepted lower fidelity initially in exchange for much higher throughput.
I would have framed AR partnerships more as reusable platform capability.
For integrations like Snap Camera Kit, we treated the work partly as a partnership initiative. If I were doing it again, I would push harder internally to frame it as a reusable AR capability layer for Flipkart, not just a one-off integration. That would likely have driven more long-term platform investment.
What I learned
This phase of my career taught me that shipping the technology is rarely the hardest part.
The hard part is making it work at scale across:
- engineering
- product
- design
- 3D content pipelines
- external SDKs
- brand teams
- device fragmentation
- and operational reality
The technology worked at Scapic with a handful of clients. Making it work for millions of Flipkart users was the actual job.
That lesson has carried forward into every AI system I've worked on since: if a system reduces uncertainty for the user and survives the messiness of production, it has a chance to matter.