Thomas Mathew

VP Engineering @ Morgan Stanley | Founding Technical Leader, DevGen.AI | Generative & Agentic AI at Enterprise Scale | 8 Granted US Patents in LLM & Agentic AI

New York City Metropolitan Area

About

For 23+ years I've engineered mission-critical systems for top-tier financial institutions and global tech leaders — UBS, S&P, eBay — across fixed income trading, market risk, regulatory reporting, index management, and high-scale e-commerce payments. Today I'm a founding engineer and technical lead of DevGen.AI, Morgan Stanley's firmwide Generative AI platform for legacy modernization and developer enablement. DevGen.AI — featured in the Wall Street Journal and Business Insider — has translated 9M+ lines of legacy code and saved ~280,000 developer hours for 15,000 developers, turning decades-old COBOL, Natural, and proprietary languages into plain-English specifications and modern code. What I work on: • Agentic AI — long-running, multi-hour research agents; multi-agent orchestration; agent memory & context management (compaction / slicing / sharing); MCP (Model Context Protocol) servers & tools; and a natural-language interface that lets non-technical users build multi-agent workflows • Hands-on GenAI engineering — LangChain, LangGraph, DeepAgents, OpenAI Agents SDK, Microsoft Semantic Kernel; AI coding assistants (Claude Code, Amp, GitHub Copilot) and authored/published reusable agent skills; LLM evaluation & observability (LangSmith); guardrails & responsible AI; on OpenAI (GPT), Claude & AWS Bedrock, with Snowflake Cortex • Full-stack delivery — end-to-end GenAI apps with Angular front-ends and Python back-ends, REST APIs, and streaming (SSE / WebSocket) • Enterprise-grade AI — on-behalf-of (OBO) authentication, AI governance, vector search, and integration with data-governance platforms and vendor agentic tools • Legacy modernization — deep COBOL / Natural program translation across program dependencies at scale I hold eight granted US patents in LLM-based code modernization and enterprise GenAI — including multi-agent orchestration and automated regulatory reporting — won the Banking Tech Awards USA 2025 (Best Use of AI/ML), and lead a globally distributed team of ~40 engineers plus a 150+ contributor inner-source community. Always happy to connect — reach out.

Experience

  • Vice President — WM & IM Technology (Architecture & Emerging Tech) · Founding Engineer, DevGen.AI at Morgan Stanley
    Oct 2022 - Present · 3 yrs 10 mos

    - Founding member & technical lead of DevGen.AI, Morgan Stanley's firmwide GenAI platform for legacy modernization (featured in WSJ & Business Insider): 9M+ lines of code processed, ~280,000 developer hours saved, 15,000 developers served. - Architected COBOL / Natural deep program-translation, traversing multi-program dependencies to auto-generate specifications — 7,000+ programs (~5M LOC) translated; pilot-to-production in ~3 weeks. - Pioneered agentic-AI capabilities: multi-hour autonomous research agents with advanced context management; a natural-language multi-agent builder; MCP (Model Context Protocol) servers/tools; and an on-behalf-of (OBO) auth pattern for enterprise-grade agents. - Delivered full-stack GenAI applications (Angular front-ends, Python back-ends) and authored & published reusable agent skills now used across teams; hands-on across LangChain, LangGraph, DeepAgents, OpenAI Agents SDK, Microsoft Semantic Kernel and AI coding assistants (Claude Code, Amp, GitHub Copilot). - Built on OpenAI (GPT), Claude & AWS Bedrock (with Snowflake Cortex) using LLM evaluation & observability (LangSmith), guardrails & responsible AI, and enterprise data governance. - Lead a globally distributed team of ~40 engineers; grew a 150+ contributor inner-source community on a 1M+ LOC codebase. - Banking Tech Awards USA 2025 — Best Use of AI/ML; MS Tech Excellence Award (2024 & 2025); 8 granted US patents. Earlier at MS (2022–2023): migrated Fixed Income products (Treasury, Agency, ABS, MBS, Commercial Paper) from legacy mainframe to a modern microservices platform; built core order-management components and Grafana-based observability.

  • Associate Director — Fixed Income Retail Trading at UBS
    Jul 2019 - Sep 2022 · 3 yrs 3 mos

    - Re-engineered legacy ServiceMix applications into a modern Spring Boot microservices architecture. - Partnered with Quant Analytics to build a PnL calculator on FINCAD analytics using proprietary algorithms. - Architected regulatory-compliance and risk-mitigation enhancements; improved performance and stability under high market-volatility load. - Automated the DevOps pipeline for a large multi-module application via the GitLab API.

  • Cognizant (New Jersey, United States)
    • Senior Architect — FI Trading, Volcker Reporting & Risk Analytics
      May 2012 - Jun 2019 · 7 yrs 2 mos

      Client: UBS Wealth Management Americas. - Built supporting trading microservices (Trade Compliance Monitor, Bond Inquiry) deployed over Hazelcast distributed infrastructure. - Implemented a price-yield calculator engine on the FINCAD F3 SDK and built a Spring-compatible wrapper around QuickFIX/J. - Delivered the FI Risk Analytics platform using Domain-Driven Design; enforced engineering standards through a shared Maven module (PMD, Checkstyle, FindBugs) and TeamCity CI. - Led an offshore team end-to-end; introduced a project-level ELK logging stack as a cost-effective Splunk alternative.

    • Architect — Standard & Poor's FI Index Management
      Jan 2010 - May 2012 · 2 yrs 5 mos

      Client: Standard & Poor's. - Re-architected the Fixed Income Index Management system with a modular, message-driven design using Spring, Hibernate, Oracle PL/SQL, Quartz, and JMS. - Drove SQL optimization and JVM performance tuning for better reliability and runtime performance.

  • Engineering Manager — Access Control & Speech Applications at Pivot Systems
    Nov 2007 - Nov 2009 · 2 yrs 1 mo

    - Built a Proximex-Pegasys P2000 access-control integration plug-in. - Delivered multiple Automated Speech Recognition applications using VXML, Nuance GSL, TellMe Studio, and BeVocal.

  • Cognizant (On-site)
    • Senior Associate / Team Lead — eBay Shopping, Payments & Platform Modernization
      Mar 2004 - Nov 2007 · 3 yrs 9 mos

      Client: eBay Inc. / eBay India. - Led Java/J2EE, XML/XSL, and JavaScript work across eBay platform modernization, Ranch frameworks, and shopping-cart experience. - Built eBay India payments capabilities including escrow workflow, reconciliation, chargebacks, refunds, bank integration, and batch processing.

    • Developer — Taxware Enterprise
      Aug 2002 - Mar 2004 · 1 yr 8 mos

      Client: Taxware Corp. - Built Java/J2EE modules for Taxware Enterprise, a global tax-calculation platform for enterprise financial/accounting systems. - Worked across rule-based tax-engine functionality, in-container JUnit testing, ClearCase MultiSite setup, SQL optimization, and performance tuning.