Director of Applied Innovation & Intelligence • Private Capital Allocator & AI Systems Architect
Forged by competitive wrestling from the age of four through his sophomore year of college as an NCAA Division I athlete, Tay developed an unrelenting work ethic, visceral grit, and a relentless drive to continually challenge limits and grow under intense friction. That physical discipline was matched by an insatiable intellectual curiosity at Brown University, where he earned a 3.6 GPA in an independent, self-directed curriculum synthesizing Philosophy, Psychology, Public Policy, and Geoeconomics to master how human behavior, institutional power, and market forces intersect. Channeling that analytical stamina and competitive drive into institutional finance, he spent years underwriting more than $700M in committed capital across 16 private equity, private credit, and venture transactions alongside direct growth-stage technology co-investments. Today, as Director of Applied Innovation & Intelligence, he converges that fiduciary rigor, behavioral insight, and builder’s grit to architect production applied AI systems, deterministic enterprise document pipelines, and frontier geoeconomic decision intelligence platforms.
Shapes vague operational bottlenecks into high-accuracy, edge-native document intake pipelines with contracted accuracy targets, orchestrating cross-stakeholder delivery and feeding legacy accounting ERPs without modifying underlying schemas.
Engineers autonomous matching platforms and deterministic decision systems from concept to production, combining taxonomy-driven weighted scoring engines, strict state verification, and operator triage controls over 500+ git commits to eliminate non-deterministic failure modes.
Develops resilient, high-throughput data and automation pipelines that ingest, reconcile, and rank disparate external records using multi-tier fuzzy matching heuristics, automated LinkedIn cross-validation, and robust batch execution workflows.
Codifies complex enterprise compliance rules into deterministic agent workflows and builds vectorless RAG routers that slash enterprise token spend by 58% with sub-200ms round-trip responses and zero-hallucination boundaries.
Launch the live multi-agent simulators and codified prompt registries directly from this dossier:
End-to-end leadership of a 5-week fixed-scope POC automating multi-document PDF invoice intake, validation, and Sage Paperless ERP routing with human-in-the-loop approval.
Built a two-sided matching platform from static mockups to production over 4 months and 500+ commits (Next.js & Supabase), featuring a taxonomy-driven weighted scoring engine and admin triage console.
Engineered and hardened a contact-matching and ranking pipeline for bidder-list automation, resolving title ambiguities, sponsor routing, and business development signals via iterative PRs.
Automated complex cross-departmental regulatory compliance, biosecurity routing, and operational decision trees across 100+ agricultural facilities.
Engineered a Generative Engine Optimization (GEO) semantic retrieval architecture and multi-engine citation auditing pipeline.
High-throughput intelligence scanning with sub-200ms latency and dynamic inference spend optimization.
Comprehensive 7-segment market study projecting 2026 and 2035 TAM matrices across commercial, industrial, and autonomous physical AI.
Keynote masterclass exploring the transformation of accounting and corporate finance leaders into multi-agent systems architects, deterministic prompt engineers, and ERP automation directors.
CPD-accredited executive briefing covering non-hallucinatory prompt design, vectorless RAG implementations, and model cost-routing middleware for CFO suites.
Comprehensive overview of corporate data isolation, zero-trust inference gateways, and regulatory compliance when deploying generative agent pipelines into legacy stacks.
True AI leadership requires knowing when not to use LLMs. Enterprise AI fails when probabilistic token predictors are misapplied to deterministic arithmetic, regulatory tables, or relational lookups. The protocol establishes an objective heuristic check to route operational workflows to their optimal compute engine.
Compound interest, LBO waterfalls, and invoice total calculations. 0ms latency, zero hallucination, and $0 API cost.
Statutory tax tables, postal codes, product SKUs, and general ledger chart of accounts. Static relational recall over probabilistic search.
Unstructured document extraction, cross-agency biosecurity synthesis, and deal-sourcing relevance wrapped with deterministic validation gates.