Work
Case Studies
Detailed accounts of enterprise platform leadership and personal AI product builds: the problems, decisions, and outcomes from systems I designed and owned end to end.
Filter case studies
12 case studies
Enterprise platform work
Standardized Enterprise Monitoring Across a Fortune 100 Infrastructure
Defined the org-wide monitoring standard and led a zero-disruption platform migration.
Read Case Study →Sole Owner of Enterprise Access Reporting Through the COVID-19 Disruption
Kept enterprise access and operational reporting uninterrupted throughout the COVID-19 remote transition, for a VPN serving 60,000 corporate users.
Read Case Study →Enterprise MFA Self-Service Platform: Web, Mobile, and QR Code Enrollment
Built three interconnected authentication platforms for the enterprise workforce: over 300,000 activation codes generated across eight years, and a self-service portal used more than 18,000 times in six languages during the COVID-19 transition.
Read Case Study →Standardized Infrastructure Monitoring Across Service Teams
Built the monitoring lifecycle platform that brought 3 million monitors under standardized management across 2,750+ application stages, and drove its adoption across service teams.
Read Case Study →Multi-Region Workflow Orchestration Platform
Own the multi-region orchestration platform: shared execution guarantees, mandatory safety checks, and self-service onboarding for development teams. Its guardrails have run 500,000 safety checks.
Read Case Study →Personal AI products
Factor-First AI Investment Platform Narrated by a Six-Persona Committee
Grew a single-ticker grader into a full investment platform: a four-factor composite (Quality, Valuation, Momentum, Health) narrated by a six-persona committee, a nightly scan of several hundred large caps, portfolio and net-worth tracking, and daily measurement of historical ranking behavior.
Read Case Study →AI Marathon Coach Grounded in 90 Days of Actual Training Data
Built a coaching platform that reasons over real Garmin training and recovery history, not generic templates, pairing a deterministic safety-first rules engine with a multi-turn AI coach and a Whoop/Oura-style data dashboard.
Read Case Study →A Fork-Your-Own Life OS an AI Agent Reasons Over, Without Owning Your Data
Designed, built, and shipped a fork-your-own life operating system: principles, vision, goals, areas, and projects as plain markdown an AI agent reasons over, under one rule the agent can read and question every file but never writes the record. A public MIT template plus a static launch site and privacy-by-construction adoption tracking, with the author's own corpus never committed or published.
Read Case Study →An LLM System Whose Best Feature Is Refusing
Built and open-sourced an LLM system whose most valuable behavior is producing no answer at all: a self-healing retrieval pipeline grounds it, a separate critic checks it, a guardrails gateway fences it, and an evaluation gate proves answer quality on every change. In the 44-question battery published with the June 2026 launch post it answered grounded questions with citations, refused advice and injections, and hallucinated zero times; on the 89-decision run of 2026-08-16 every answer-or-decline decision was correct and one answer was unfaithful to its sources.
Read Case Study →Context Architecture Beats Documentation Dumps
Designed a context-partitioning method for AI agents that replaced documentation dumps with scoped spec slices, applied at enterprise platform scale, to a training coach, and to an investment committee. The clearest demonstration on record: two full-brief dispatches stalled with no files; the same task with a halved brief and a named first file passed in one shot.
Read Case Study →A One-Person Engineering Platform Run Through Spec-Driven Development
Built and operate, as one person, a spec-driven engineering platform: a frontier-tier session plans and reviews, cheaper executors implement from written briefs against a work queue of roughly 1,500 items, changes land through a locally attested gate whose checks are tripwired weekly, and a scheduled pass edits the platform's own instructions. From June 10 through August 2026: 4,141 commits and 2,244 merged pull requests across 25 repositories, with the failures kept on the record.
Read Case Study →A Self-Governing Node in a Solo Fleet
Turned an always-on ARM board into a self-governing node: a resident agent that reads a written constitution first on every run, acts only through a code-enforced fence, defers anything privileged to a phone-approved queue, and files its own backlog every night. Around it, the security floor and the local CI mirror a one-person fleet needs and cannot staff.
Read Case Study →