Rohan Sanil and the Things a Receipt Cannot Tell You

A cash register is an excellent witness with a very narrow field of vision. It can tell you what sold, when, and for how much. Ask about the person who walked in, considered a shelf, waited for help and left without buying, and the register has remarkably little to say. Rohan Sanil made that silence a business proposition.

Sanil co-founded Deep North in 2016 with Jinjun Wang. The company, now called StrataVision, applied computer vision to the physical places where customers move, hesitate and queue. His career offers a useful way to approach artificial intelligence: through the things a shopkeeper would like to know before closing time. The questions are pleasantly unglamorous. Where did people go? How long did they wait? Did anyone help?

The technology has changed, and so have the company names. Sanil has kept working with video, products and the commercial task of making them useful. Today he is a co-founder of Bluebird AI. To understand that next chapter, it helps to follow an earlier piece of work that had little to do with shopping baskets.

A highlight reel before a checkout lane

On October 1, 2019, a US patent for a system for automatically editing video was granted with Sanil and Wang named as inventors. Its subject was the construction of a highlight video from a longer media file. The system extracts frames, represents their features, groups related material and selects segments to combine.

Anyone who has returned from a trip with too much footage can appreciate the problem. A recording preserves duration generously; an audience is usually less generous. Selecting the useful moments requires some way of deciding which parts deserve attention. In this invention, analysis of the images helps make that selection.

The patent gives Sanil’s career a technical anchor. He appears in the record as a co-inventor alongside the collaborator with whom he founded Deep North. Before the retail dashboards and corporate partnerships, here was a concrete question about moving pictures: how do you find the parts that matter? The later business would ask a related question in a very different room.

Several companies before the familiar one

Deep North arrived after earlier rounds of company-building. Sanil founded Tri-Cad, Atstream Networks and Akiira Media Systems. Those ventures placed him in the overlapping work of developing products, pursuing business and raising capital. A founder’s job description can look suspiciously like the contents of an entire office.

His professional timeline places Tri-Cad and Infoglobe in the 1990s, Atstream around the turn of the century, and Akiira beginning in 2008. He also worked in business development at Cambridge Solutions and in product management at MetricStream. These were different vantage points on a recurring commercial challenge: getting a technical idea into an organization that has reasons to buy it.

His education combined a bachelor’s degree in mechanical engineering at Karnatak University with a master’s in management science at the University of Dayton, where he studied from 1992 to 1994. The combination sits comfortably beside a career that crosses product and business responsibilities. It supplies useful background without needing to explain every subsequent decision.

There is a more personal glimpse in a public recommendation from Anish Bole, who calls Sanil an “eternal optimist.” Bole describes persistence in pursuing business and a mentoring style that made colleagues comfortable seeking help. The description is a colleague’s assessment, and it brings a human scale to the succession of executive titles.

The customer the till forgot

At Deep North, the proposition centered on making existing camera infrastructure useful for business analysis. In March 2020, the company announced a $25.7 million Series A led by Celeres Investments, with Engage and AI List Capital participating. Plans included expanding the team, developing the platform and exploring further markets.

The commercial argument was about an information imbalance. Online businesses could examine a visitor’s path through a website. A physical retailer needed ways to understand a journey through an actual building. Deep North offered analysis of video already being captured in stores, shopping centers and other properties. The camera installation became the starting point for another kind of business tool.

In October 2021 came a $16.7 million Series A-1, co-led by Celesta Capital and Yobi Partners, with Conviction Investment Partners participating. The announced purpose was to add key hires and expand services. The company described uses including matching staffing to customer traffic and understanding which departments attracted attention.

Those uses give the money some context. A funding announcement is a milestone in a company’s financing; the operator still has a shift to schedule. Sanil’s pitch connected the two scales. Investment supported a product whose everyday purpose was to help someone make a more informed choice about a store.

By November 2020, Sanil was also taking part in the public conversation about offline retail’s digitization. A Retail Refined live event hosted by Melissa Gonzalez paired him with Radius8 founder Sandeep Bhanote and consultant Ken Pilot. Its subjects included location-based marketing, analytics and traffic. The company-building took place alongside the work of explaining the category to people running stores.

A bakery in the quiet corner

Sanil’s writing about shopping centers supplied a particularly agreeable illustration of his thinking. In a 2022 article, he suggested placing a pop-up bakery in a low-traffic area and testing whether it encouraged people to circulate. This was a proposed experiment, rather than a reported customer result. It also had the advantage of being understandable before the coffee arrived.

The broader argument covered tenant mix, lease negotiations and the effectiveness of events. A property owner could examine where visitors went and how long they stayed. An event could be assessed for the traffic it brought to an area and the effect on neighboring tenants. The analytical interest extended beyond the boundary of one store.

Elsewhere, he made the case for better visibility into the in-store experience: browsing, queues, merchandising and access to assistance. In those arguments, computer vision served a familiar retail ambition. Customers should find what they want with less friction, and managers should understand where the experience could improve.

His contributed writing on applications of AI returned to labor planning, operations and customer journeys. That repetition tells us something about his professional focus. He kept explaining the technology through decisions a business already recognized. A manager did not need to acquire an entirely new set of concerns to understand why the product might be useful.

When two counts disagree

The checkout lane narrowed that broad ambition to a specific comparison. In his writing on loss prevention, Sanil described connecting camera-generated item counts with point-of-sale counts. A discrepancy could produce an alert for staff to review before the transaction finished. The possible causes included erroneous billing as well as theft.

The human handling mattered. He described giving customers a chance to pay for an unscanned item without beginning with an accusation. An alert creates a question for an associate. What happened during the transaction? The useful outcome depends partly on how the employee approaches the answer.

On April 1, 2025, US patent 12,266,246 was granted for self-checkout verification systems and methods. It names Sanil, Abhijit Deshpande and Wang as inventors. The described approach compares item counts and categories derived from camera images with those in a point-of-sale record, and can generate an alert when they differ.

The difference from the earlier editing patent is striking. One selects scenes for a video; the other checks accounts of a transaction. Both give analysis of images a defined task. That makes the patents a useful thread through a career otherwise spread across ventures, responsibilities and changing company names.

The next camera, the same practical question

Partnerships were another part of the work. In September 2022, Deep North and Dell Technologies announced a collaboration combining Dell edge computing with video analytics. They planned a joint demonstration at Groceryshop in Las Vegas. The application list included queues, staffing, shelf engagement and checkout information.

Sanil’s founding chapter at Deep North also needs to be separated from the company’s later leadership. In July 2024, the business announced its new StrataVision name under CEO Bjoern Petersen. The rebranding accompanied a new leadership team and a broader growth strategy. Sanil remains part of its founding history.

His subsequent Bluebird AI chapter carries the interest in video forward. The company’s public product descriptions combine camera hardware, AI and business intelligence. His own announcement of joining described a co-founder and product role. Another venture has provided another setting for the technical and commercial work that runs through his history.

“how can I implement AI to solve my defined problem?”

Rohan Sanil · 2022

There is a sensible restraint in the question Sanil has offered to aspiring AI executives. The order matters. First identify what needs fixing. Then choose how to address it. For a career involving so many cameras, it is an appropriately grounded place to end: with the person who needs an answer, standing somewhere just outside the receipt.