Tazo Lezhava

Global agentic commerce volume is projected to reach up to $5 trillion by 2030, yet less than 1% of enterprise e-commerce infrastructure currently meets the minimum standards to be recommended by Large Language Models. In this session, Tazo Lezhava dissects the critical infrastructure bottleneck plaguing modern retail: the fatal reliance on rigid, legacy lexical search.

This talk will break down why traditional keyword matching is fundamentally incapable of communicating with the next generation of autonomous AI shopping agents, resulting in high zero-result rates and lost revenue. Attendees will explore the technical architecture required to bridge this gap, specifically focusing on how to deploy Business-to-Agent (B2A) sidecar APIs without executing massive, disruptive database overhauls.

The session will dive into the mechanics of utilizing custom multi-modal vector embeddings and dynamic KV caching to process complex visual and semantic search intents. Furthermore, it will reveal how this exact same B2A infrastructure can be utilized to automate checkout validation, actively slashing margin-destroying return rates in highly subjective categories like fashion and personal care. Join this session to learn how to transition away from text-based optimization and build the true multi-modal data protocols required to capture the imminent wave of AI-driven commerce.

Tazo Lezhava is the CEO and Co-Founder of Quissly, a deep-tech company building foundational Business-to-Agent (B2A) AI infrastructure for enterprise e-commerce. Prior to founding Quissly, Tazo built enterprise-grade data models as a quantitative analyst at BlackRock in San Francisco, designing AI architecture to manage over $6.5 billion in AUM, and led scalable data science initiatives at TBC Bank.
Backed by 500 Global and an alumnus of Columbia University, Tazo is currently leading Quissly’s deployment of multi-modal sidecar APIs across enterprise retail. His work focuses on bridging highly complex quantitative AI systems with commercial execution, actively transitioning global retailers from legacy keyword search to semantic discovery to future-proof their catalogs for the autonomous agent economy.