
▌
Ethan Smith
Full-Stack Software Engineer · AI Platform Engineering
Decision Intelligence @ Workday

// about
Senior software engineer with 5+ years at Workday building AI-powered planning tools and full-stack applications. Currently in Workday's Agent Factory, the company's dedicated AI organization, working on Decision Intelligence, a new product line and an AI-native platform where finance teams explore and model their data in plain language across governed sources like Adaptive Planning, Prism, Snowflake, and uploaded spreadsheets.
Before that, I spent three years modernizing legacy systems, migrating Tapestry/JS pages to React + REST APIs, building report schedulers, and implementing time hierarchy features across the full stack. I studied Computer Science and Business Administration at USC. Outside of work, I'm into backpacking, Broadway, vinyl collecting, baking, and strategy games like Civ and Catan.
// lately
Reading
The 7½ Deaths of Evelyn Hardcastle
Playing
Factorio: Space Age
Listening
Trying Times — James Blake
// experience
Senior Software Development Engineer
2021 — presentWorkday · Adaptive Planning → Agent Factory
Spent my first five years in Adaptive Planning, then moved into Agent Factory, Workday's dedicated AI organization. It builds autonomous agents as new product lines of their own, rather than AI features added onto existing products.
Agent Factory · Decision Intelligence
Workday Decision Intelligence is a new product line, not an add-on to an existing product: an AI-native platform where finance teams ask questions, model scenarios, and commit decisions across governed data sources including Adaptive Planning, Prism, Snowflake, and uploaded spreadsheets. Customers buy it in its own right. I connected Adaptive Planning reports directly into the product, so users no longer export a report and re-upload it as a file to analyze it, and I contributed to the agent skills that integrate Adaptive Planning on both sides: hypercube data queries for reads, and data write-backs to planning sheets for writes. I also own performance testing and end-to-end tracing across services and LangSmith.
Adaptive Planning · Planning Agent
On the team that built the Planning Agent for Adaptive Planning, a LangGraph agent that shipped to GA in Mar 2026 and now has 1,300+ customer accounts with recorded usage (Aug 2026). I contributed to its data exploration, variance analysis, scenario planning, and contextual help skills (all four gated at a 95% eval pass before release), designed the MCP client-server architecture with progressive tool disclosure and subagents to cut context bloat, and built the elicitation framework that rendered planning widgets inline within chat.
Adaptive Planning · Full-Stack Engineering
Built a portable report viewer that ships as a widget across multiple applications, with 544 instances and 12,760+ reports added in its first three months. Also replaced a manual report workflow with a cron-based scheduler for snapshot generation and notifications, used across 6,500+ enterprise customers, and migrated legacy Tapestry/JS pages to React with REST API backends.
Software Engineering Intern
2020Xactly
Automated SOC 2 audit evidence collection across Jira, Azure DevOps, Workday, and Salesforce.
Product Marketing Intern
2019Teradata
Product marketing for enterprise data analytics and cloud data warehousing.
Data Engineering Intern
2018Warner Bros. Digital Networks
Data engineering for Warner Bros.' direct-to-consumer digital platforms.
// projects
Vacation Price Tracker
Tracks flight and hotel prices for planned vacations with automated daily checks, historical trend charts, and threshold alerts. FastAPI backend with Temporal workflows and an LLM chat assistant that searches flights and hotels through MCP tools; Next.js web and Expo mobile clients; multi-provider flight search with automatic failover.
Showbook
Personal entertainment tracker for live shows — concerts, theatre, comedy, and festivals. Track your show history, discover upcoming events, and see your patterns.
MCP Review
AI-powered code review CLI for solo devs who push to main without PRs. Spawns 4 MCP tool servers over stdio (git diff, file context, convention scanner, related files) and lets the LLM drive its own context gathering. Reviews reference actual project patterns and lint configs, not generic advice.
MCP Budget Governor
Cost-denominated limits for MCP servers: per-user quotas, per-tool limits, and a global spend circuit breaker. Python and TypeScript packages share one atomic Redis contract, so services in both runtimes enforce the same budget.
MCP Queue Doctor
MCP server that diagnoses pg-boss and graphile-worker queues. It turns retry storms, stuck jobs, missed schedules, and expiry overruns into evidence-backed findings with the safest recovery steps.
Provider Router
Routes one capability across interchangeable providers with ordered failover, normalized results, and explicit handling for degraded responses. Zero-dependency packages on PyPI and npm share one set of behavioral test vectors, so the two runtimes cannot drift apart. In use for flight-provider failover and venue geocoding.
// contact
Get in touch
Let's talk about AI agents, Factorio base designs, or whatever you're working on.