Lavan
Sornarajah
With a background grounded in neuroscience, I have always viewed technology through a scientific lens, looking for the signal in the noise.
I have always been an engineer of intelligence; the medium just shifted from wetware to software.
Side quests and rabbit holes. Tools built for curiosity, utility, and the occasional deep dive.
3D knowledge graph of the adtech agentic stack. Control plane, data plane, execution plane.
AI-powered bookshelf scanner. Detect spines, enrich metadata, explore your library as vectors.
CRT terminal sequence you probably shouldn't run. Trust issues optional.
The "Sigma" Strategy
Recruited to deploy the UK Go to Market strategy for Sigma, MiQ's AI powered unified intelligence platform spanning TV, Commerce, and Context.
- Autonomous Intelligence: Architected a FastAPI microservice to orchestrate 30+ parallel API calls. Engineered a custom "Domain Specific Language" (DSL) to harmonise fragmented ACR and YouTube signals into a unified schema and agentic analysis. Resulting in slashing strategy generation time from 4 hours to 2 minutes.
- Cortext: Applied Attentional Gating and Dual Coding neuroscience principles to build a Multimodal Analysis Engine. Used Google Gemini & Meta's Segment Anything Model (SAM) to measure brand & product pixel density and combined with emotive triggers to produce pre-campaign creative scores.
- Measurement Planning: Engineered a conversational agent that acts as a "Complexity Abstraction Layer." Users upload media plans/RFPs, and the agent parses variables to recommend optimal solutions, effectively "solving" the measurement type, and vendor capability matrix in minutes.
Gold Standard & "NORI"
Leading the industry through the post-cookie transition while building internal automation tools that secured official endorsement from the UK Government (GCS).
- Compliance Agent: Built a "Core + Composable" agentic platform using OpenAI multimodal models to ingest diverse evidence types ([app-]ads.txt, bidstream logs, sellers.json etc.) along with tool access to validation scripts, TCF GVL and TAG registries. Reduced audit times by 95%, creating a fully auditable, Human in the Loop supply chain system.
- Enterprise RAG (NORI): Architected an enterprise knowledge retrieval system using Agentic routing to query complex datasets from technical industry standards (IAB Tech Lab, Prebid, UID).
- Synthetic Alignment: Engineered a pipeline generating synthetic Q&A pairs during document and codebase indexing. Ensures AI systems understands intent ("How do I fix error X?") rather than just semantic keyword matching. Built custom retriever using Parent-Child chunk strategy for vector search to ensure high data generation fidelity.
The Attention Economy
Instrumental in moving the industry from "Verification" (Did it load?) to "Attention" (Did they look?). Commercialised programmatic products bridging biological eye-tracking data with programmatic buying logic (demand) and curated inventory (supply).
- Commercial Impact: Secured two Innovate UK grants by co-pitching the fusion of AI and behavioural science.
- Validation: Ran successful pilots for GSK and Haleon, proving that buying for Attention drives measurable Brand and Sales Lift over proxy metrics.
- Product Solutions: Developed automated workflows for custom bidding algorithms for DSPs and high attention inventory curation data sets for supply platforms.
The Exit & Investment
Bootstrapped a managed service trading desk from zero to profitability and strategic exit after 9 years.
- The Build: Developed custom algorithms and advanced ID management solutions for multi-DSP orchestration across The Trade Desk, DV360, and MediaMath.
- The Pivot: Orchestrated a profitable exit and leveraged capital to become the largest external investor in Avocet Systems (Lumen), correctly identifying Attention Metrics as the next major industry currency.