Kyle Thomas - Resume

PROFESSIONAL SUMMARY

Product leader with 15+ years across platform, data, and automation, now leading agentic AI product work at JPMorgan Chase, including multi-agent orchestration and MCP integration. Manages a team of 8 and has delivered $6M+ in annual impact. Pairs hands-on building with governance-first design for regulated environments.

SKILLS

Agentic AI: Multi-agent orchestration (orchestrator and supervisor patterns, agents-as-tools), Model Context Protocol (MCP), RAG, LangGraph, AWS Bedrock, AWS Strands Agents and AgentCore, event-driven architecture

AI Evaluation & Governance: Golden datasets, LLM-as-judge, confidence gating, human-in-the-loop workflow design, agent risk controls (prompt injection, tool permissions, human approval gates)

Product & Platform Leadership: 0-to-1 product development, API-first architecture, self-service tooling, multi-year roadmaps, OKRs and success metrics, spec-driven development

Technical: Python, SQL, Claude Code, VS Code, GitHub Copilot, Jira MCP, n8n, Docker, GitHub, Alteryx, integration architecture

PERSONAL PROJECTS

AI Sports Analytics Platform. Confidence-gated prediction engine that suppresses low-certainty recommendations.

Personal AI Assistant. Multi-model (Claude API, Gemini fallback) with workflow orchestration, calendar, and IoT automation.

Algorithmic Trading System. Phased validation gates (backtest, paper, live) before advancing real capital.

CERTIFICATIONS

AWS Certified AI Practitioner (AIF-C01)

eCornell Agentic AI Architecture Certificate

EDUCATION

University of Denver

MBA (2015)

Illinois State University

Bachelor's, Finance (2010)

EXPERIENCE

JPMorgan Chase - Chicago, IL

Senior Product Manager, Product Lead | Oct 2021 – Present

Led an agentic AI pilot for client servicing, with human-in-the-loop gates under a "human always sends" principle.
Lead a classification and extraction model that routes inbound client-servicing requests to agentic automation or manual handling.
Defined the evaluation methodology (golden datasets, LLM-as-judge scoring) used to measure agent quality before scaling.
Drive MCP integration and multi-agent orchestration for the platform.
Defined strategy for a GenAI rules optimization system that detects duplicate and conflicting decision logic across thousands of product configurations, routing low-confidence results to expert review; targeting 40% less manual maintenance.
Reduced onboarding rejections by 80% (25% to under 5%) within the reviewed product subset with an eligibility framework that blocks ineligible product combinations upfront, preventing 3,500+ rejected applications annually.
Scaled a self-service product data platform from 0 to 400+ users, unlocking 1,200+ annual product changes previously blocked by $50K cost and 6-month engineering cycles, worth $4M+ in annual cost avoidance.
Reduced enterprise client onboarding time by 65% (20+ days to under 7) by modernizing API infrastructure powering 18,000+ annual onboarding requests.
Achieved zero critical audit findings over 4 consecutive years with a risk-based change framework: automated approvals for low-risk updates, controlled release trains for high-risk changes.
Retired a 12-person manual billing function ($1.5M annual savings) by digitizing offline spreadsheet processes and prepopulating data flows across platforms.
Grew from IC to product lead of 8 (5 hires, sub-team led by a Principal PM); own roadmap, OKRs, headcount, and executive QBRs.

Lumen Technologies - Broomfield, CO

(formerly CenturyLink)

Director of Tax Planning & Transformation | Aug 2015 – Oct 2021

Promoted twice in 4 years from Tax Manager to Director, applying indirect tax expertise (sales/use, VAT, GST) to automation.
Built the business case for and founded a 0-to-1 automation platform team of 3, owning its charter and roadmap.
Shipped 20+ self-service applications and trained non-technical analysts to build their own, removing engineering dependencies.