ABOUT KETUPA
AI shopping platforms provide powerful agents, data connections, workflows, analytics, and implementation controls. But every merchant still has its own product truth, category complexity, customer needs, business rules, priorities, and exceptions.
KETUPA provides the merchant- and category-specific delivery required to turn those platform capabilities into reliable real-world shopping experiences. We work within the agent and commerce stack already in place—establishing the required shopping behavior, testing complete journeys, implementing accepted optimizations, verifying the results, and maintaining critical behavior over time.
LEADERSHIP
Founder & CEO
AI shopping agent behavior · Product strategy · Commerce
Bufan leads KETUPA’s company vision, product direction, and work on AI shopping agent behavior.
His work focuses on how agents interpret shopper needs, apply product and business requirements, qualify uncertainty, recommend and rank products, and adapt as the shopping journey changes.
At KETUPA, he turns that work into merchant-specific shopping agent optimization and reliability programs.
Co-Founder & CTO
Former Microsoft Partner Engineering Manager · Principal Engineering & Applied Science Leader · Tsinghua CS PhD
Jian-Tao leads KETUPA’s technology strategy, architecture, engineering, and productization.
His experience spans search and ranking, information retrieval, vector and hybrid search, RAG, knowledge systems, conversational AI, and agentic search.
At KETUPA, he is building the technical system that turns merchant- and category-specific shopping agent behavior into scalable evaluation, implementation, and reliability workflows.
Founding Member · Applied AI & Systems
Amazon Production Systems · Applied AI · Distributed Systems
Varshith builds the production and automation systems behind KETUPA.
His experience spans Amazon production systems, distributed systems, applied AI, agentic workflows, RAG, evaluation, integrations, and production reliability.
At KETUPA, he develops the tooling and infrastructure that make shopping agent optimization more repeatable and scalable.
BEYOND THE FOUNDING TEAM
WHY KETUPA EXISTS
The hard part is often not whether the agent can retrieve the right fact.
It is whether the right shopper requirement, product condition, merchant rule, or exception actually changes what the agent recommends, ranks, claims, asks, shows, and does next.
That behavior can work in one response and break under a different shopper wording, product card, follow-up, later turn, or system change.
KETUPA exists to make merchant- and category-specific shopping agent behavior explicit, implementable, measurable, and reliable.
WHAT KETUPA IS BUILDING
KETUPA is building a repeatable system that turns merchant and category requirements into implemented and verified agent behavior. It connects business-approved behavior, category-specific evaluation, targeted implementation, before-and-after verification, and retained regression coverage—working through the agent and commerce stack already in place.
Built for merchants and AI commerce platforms that need important shopping journeys to work correctly for their specific products, categories, policies, and customers.
FOR PLATFORM PARTNERS
KETUPA helps platform customers translate strong platform capabilities into shopping experiences that work for their specific products, categories, policies, and customers.
We can support merchant onboarding, category-specific configuration, implementation, optimization, validation, and ongoing reliability. KETUPA can work alongside your delivery team, co-deliver with your agency partners, or lead the merchant-specific optimization workstream.
We work through the capabilities and controls already available in your platform. KETUPA does not replace the agent platform—we help joint customers realize more value from it.
Explore a Delivery PartnershipSTART WITH ONE IMPORTANT SHOPPING JOURNEY
Bring one agent, one category or product family, and one shopping journey that matters. KETUPA will scope a focused first engagement around the behavior that matters most—without replacing the agent or platform already in place.
