Beyond Tag and Map: Building Long-Term Memory Twins with LLMs and Real-Time Telemetry Graphs
Commercial real estate is a massive global asset class, but its operational performance is bottlenecked by the "Implementation Gap." Modern buildings generate petabytes of telemetry, yet 90% remains "dark data"—cryptic, unstructured legacy shorthand trapped in silos. Traditionally, deploying intelligent analytics required months of manual tagging and mapping, killing project ROI before it began.
This session details a new approach: replacing the manual data-tagging bottleneck with autonomous, agentic processors. We will walk through the architecture of a Memory Twin—a system that fuses real-time stream data, metadata, and static building documentation (such as manuals and historical logs) into a unified semantic graph. We will explore how to ground LLMs in building physics to prevent hallucinations, how non-deterministic agents reason over mechanical anomalies (distinguishing sensor drift from physical valve failures), and how to capture the transient diagnostics of on-site engineers into a persistent, self-improving operational memory.
Session Speaker
Mark Sorsa-Leslie is a tech entrepreneur, Chartered Surveyor, and venture builder dedicated to solving complex real estate challenges with practical emerging technology. As Co-Founder of Auttaa AI, he drives the development of domain-specific AI agents engineered to automate real estate workflows, reduce administrative overhead, and boost workforce productivity. Concurrently, as Co-Founder and Co-CEO of Beringar, he steers the product strategy and growth of an IoT platform that uses real-time ambient data to optimise building energy usage, space efficiency, and overall indoor health.
With over three decades of experience bridging physical real estate with digital innovation, Mark holds a Chartered Surveyor qualification (MRICS) and an MBA from Middlesex University. His entrepreneurial track record includes co-founding tech companies like Open Era and Remedios, as well as leading key software initiatives at Nordic IT provider Tieto. Across his career, Mark’s focus remains centered on taking complex tools—from sensor networks to tailored AI agents—and turning them into measurable, real-world operational performance.
