Building world-class systems at scale — from autonomous AI agent frameworks to mobile platforms serving 2 billion users.
I'm a Staff Software Engineer with a decade of experience building production systems that operate at global scale. My work spans Android platform engineering, autonomous AI tooling, and distributed infrastructure — with a consistent track record of shipping complex, high-stakes programs on time.
I led technical programs at Meta / Instagram on a 2B+ user platform, built crash detection infrastructure saving an estimated $50M per year, and technical-led the feature that drove 180M new user acquisitions in a single quarter.
My recent engineering focus spans autonomous AI developer tooling — multi-agent systems that coordinate, verify each other's work, and run reliably without constant supervision — alongside deep Android platform instrumentation and cloud-scale mobile test infrastructure.
"The best engineers don't just write code — they build systems that outlive them."
From on-device instrumentation to autonomous AI systems and cloud infrastructure.
From Disney to Meta to autonomous AI systems — every stop shipped something that matters.
Shipped systems with real, measurable impact — from autonomous AI agents to billion-user mobile platforms.
A full AI fitness coach, shipped end-to-end and live on the App Store: chat-first coaching, photo-based meal logging with instant calorie estimation, and a 0–100 Physique Score that scans your progress and hands you the next step. Built on SwiftUI + Supabase with a Gemini-powered coaching engine, StoreKit billing, and an automated QA + self-improvement loop. The v3.1 release adds a native Apple Watch app — Today protocol, quick-log, Physique Score, and a watch-face complication — bundled right inside the iOS app.
Autonomous multi-agent coding framework with a four-tier hierarchy: Oracle → Manager → Lead → Coder. Each layer operates with phase-gated execution — agents can't advance without presenting evidence of completion. A SQLite-backed code knowledge graph over the entire Python source tree lets agents reason about cross-file dependencies and promote learning across runs. Resumable workflows survive session failures. Resolved three classes of deep reliability bugs: runaway self-recursion in agent spawning, inter-agent communication deadlocks, and venv bootstrap failures — making fully autonomous, unsupervised coding runs reliably production-safe.
Zoomable production map of a large-scale mail application — 134 screens, 220 navigation edges — derived from instrumentation source code rather than hand-drawn diagrams. Won an internal A/B test against the legacy tool and shipped to production. Includes static-analysis parsers (Compose wireframe extraction, ScreenGraph parsing, Android manifest deep-link resolution, Flux navigation extraction) and a live emulator capture engine that drives the UI over ADB, self-heals from ANRs, and verifies captures against element inventories.
Technical lead for Trial Reels on a 25+ engineer cross-functional team at Instagram. Launched Q1 2024 — drove 180M new user acquisitions in a single quarter. Simultaneously led the platform-wide migration of 15M+ lines of code to Kotlin + Jetpack Compose across the entire 2B+ user application — crashes down 42%, startup time up 35%. Built the coroutine networking framework that was adopted by 8 product teams, processing 10K+ requests per second at 99.99% uptime with 28% latency reduction. Mentored 12 senior engineers through the migration.
Crash detection infrastructure processing 500M+ daily events across the Instagram Android app. Reduced mean time to detect (MTTD) from 6 hours to 15 minutes. Estimated annual engineering cost savings of $50M through faster incident response. ML-powered crash analysis surfacing patterns across 40+ product teams.
Unified super-app consolidating 3 separate legacy fitness apps (12M combined users) into one. Infrastructure redesign cut costs by $2.3M/year. AWS backend supporting 50K+ concurrent users at 99.95% uptime with <100ms p95 latency for 8M+ DAU.
Containerized Android emulator running on nested-virtualization GCE instances, driven over an IAP tunnel. Serial-agnostic capture pipeline for automated UI testing at cloud scale — enables headless Android test execution at Google Cloud speed without physical devices.
Built a full AI-powered chat application and automation layer on top of the ChatGPT API. Includes a custom interface, persistent conversation context, intelligent response routing, and automated workflow triggers — designed for real-world deployment with session handling, streaming responses, and extensible tool integrations. Shipped end-to-end: backend API, frontend UI, and cloud deployment.
Designed and shipped a full production website for an independent client. Handled the complete engagement — requirements, custom design, responsive development, and live deployment. Delivered on time with a clean handoff and zero post-launch issues.
10+ years building production systems at scale — from 2B-user mobile platforms to autonomous AI developer tooling. Always interested in connecting with great engineers and ambitious teams.