Jai Chopra

Jai Chopra is an applied-AI product leader whose career spans assistive wearables, autonomous vehicles, vector databases, and multimodal consumer products. As a product manager on Uber’s computer-vision team in 2026, he helped develop closed-loop evaluation for Uber Eats food photography, balancing image quality against merchant identity, customer trust, and production economics.

Chopra studied computer science and economics at the University of Sydney and earned a master’s degree in computer science from the University of New South Wales. He co-founded Edisse, a Sydney wearable-technology startup focused on helping older adults live independently, applying machine learning and simulated training data to fall detection. After consulting at Deloitte, he became chief technology officer of menswear styling company Kent & Lime, where he developed personalization and recommendation systems.

At Cruise, he worked on machine-learning infrastructure, data platforms, and simulation for autonomous vehicles. He subsequently became head of product at LanceDB, a vector database for multimodal AI applications, where his work included making computer-vision datasets easier to inspect and validate. His professional history and LanceDB’s collaboration with FiftyOne chart his movement from building individual AI features toward the infrastructure supporting them.

• Authenticity as a product requirement. For Uber Eats, better food photography must preserve the actual dish, portion, and merchant identity. Added shrimp, missing sauce, implausible plating, and homogenized restaurant imagery are product failures even when the image looks convincing.

• Selective multimodal routing. Photographs, descriptions, and metadata determine whether an image should be enhanced or left untouched. Editing unnecessarily increases compute costs and can degrade strong originals; different routing decisions can also balance model quality, latency, and expense.

• Layered safeguards and reward hacking. Generation passes through repeated checks for faithfulness, completeness, realism, and policy compliance before publication. Chopra highlights reward hacking in image generation, where superficial changes satisfy narrow evaluation criteria without creating meaningful improvements.

• Observability tied to marketplace outcomes. Structured orchestration logs make failures understandable across engineering, design, and product teams. Merchant feedback, internal testing, benchmarking, and rollback feed a production feedback loop, while cart additions and completed orders reveal whether improvements actually help customers across different cuisines, regions, and devices.

Chopra developed these ideas publicly in a joint presentation with Uber colleague Soumya Gupta. He has also expressed support for open technology and protocols.