Staff Software Engineer — Agentic Systems & AI Infrastructure

🎯 Executive Summary

12+ years of industrial software engineering across iOS, Android, web, and backend at 3+ billion MAU scale — now building production AI agent systems at Meta. 2+ years operating multi-agent architectures with self-improving feedback loops, schema-constrained handoffs, and agent infrastructure (versioning, CI reliability). Rated Greatly Exceeds Expectations; ranked top 3 of 700+ engineers org-wide in coding contribution, driven substantially by AI-assisted development workflows.

Key Achievements:

• 🤖 Built a self-evolving 3-agent Swift migration system (converter → brain → feeder learning loop)
• 💰 Shipped a codemod agent saving $5M+/yr; an AI exploration agent driving 0.01%+ revenue lift
• 🏆 Technical Excellence Award winner at Snap Inc. (company-wide recognition)
• 📊 Top-10 all-time committer to the Snapchat iOS codebase; Facebook Monetization Swift contribution ranked #1 in 2025
• 🧪 Led 150+ A/B test launches (up to 200k LOC in a single test)
• 👥 Mentored 5+ engineers to E5 promotion as tech lead

🤖 Agent & AI Systems Expertise

• Multi-agent architecture & orchestration: task-specialized subagent pipelines, schema-constrained handoffs, self-improving agent loops, node/graph-based scaling architectures; agent infrastructure (marketplace/plugin versioning, CI reliability tooling)
• Applied agents: combinatorial UI exploration agents, daily codemods (deadcode deletion, schema-sync), large-scale code migration agents
• Familiar with RAG architectures and retrieval-augmented pipelines
• Languages: Swift, Objective-C, TypeScript, PHP (Meta variant); OOP, protocol-oriented, functional paradigms
• Tooling: Cursor, Xcode, VSCode, GitHub, livegrep, macOS terminal/bash; SQLite and large-scale SQL querying in GCP
• iOS depth: Core Bluetooth, GCD, Core Data, Swift Concurrency, UIKit, MapKit, APNS, Core Location

💻 Technical Skills

Languages & Paradigms

• Expert: Swift, Objective-C
• Proficient: TypeScript, PHP (Meta variant), Python, Shell scripting
• Paradigms: OOP, Protocol-Oriented Programming, Functional Programming

Agent & AI Systems

• Architecture: Multi-agent pipelines, schema-constrained handoffs, self-improving feedback loops, node/graph-based scaling
• Infra: Agent marketplace versioning, plugin/dependency management, CI reliability tooling, deterministic + LLM-judgment hybrid flows
• Familiar: RAG architectures, retrieval-augmented pipelines

iOS Frameworks & Technologies

• Core: UIKit, SwiftUI, Swift Concurrency, Combine, Core Data, Core Animation
• Media: AVFoundation, Photos, Camera, Core Image
• System: GCD, Core Bluetooth, Core Location, APNS, Push Notifications, MapKit
• Tools: Xcode, Instruments, LLDB, CI/CD pipelines

Engineering Practices

• Experimentation: 150+ A/B test launches, metrics-driven development, SQL analytics
• Architecture: Modular architecture, large-scale refactoring, dependency injection, clean architecture
• Leadership: Cross-team initiative planning, incident response (P0/P1), engineer mentorship

💼 Professional Experience

Meta Platforms, Inc. | June 2025 - Present

Staff Software Engineer, Tech Lead — Facebook Monetization (Facebook / Instagram)

Ramped extremely fast — Exceeds in H1 PSC, Greatly Exceeds in H2 PSC; ranked top 3 of 700+ engineers org-wide in coding contribution.

Agent Architecture & Infrastructure:

• 🤖 Built a self-evolving 3-agent Swift migration system: Agent 1 (Converter) performs code conversion and emits key metrics; Agent 2 (Brain) consumes those metrics to open PRs improving Agent 1’s prompts, skills, and MCPs; Agent 3 (Feeder) surfaces additional signals to trigger further learning
• 🔗 Designed a 3-layer subagent pipeline for mobile UI verification — spec compiler, navigator, comparator — passing schematized data between layers to produce structured verdicts
• 🕸️ Pioneered a node/graph-based agent architecture for scaling, replacing traditional E2E test architecture entirely; now supports 100+ engineers’ workflows
• 📦 Contributed to core agent marketplace infrastructure: versioning system and plugin/dependency version management
• ⚖️ Architected agentic flows mixing deterministic Python scripting with non-deterministic LLM judgment, continuously tuning the stability-vs-scalability tradeoff

Applied Agent Impact:

• 💰 Built Craft Scout, an AI exploration agent testing ad rendering across combinatorial dimensions (ad style × ad ID × device settings) — drove a 0.01%+ revenue lift for Facebook
• 🧹 Built a daily-run codemod agent for deadcode deletion and schema-sync cleanup, saving $5M+/year
• 🔄 Led an Objective-C → Swift migration codemod covering 30k+ LOC in 3 months
• ✅ Contributed to company-level CI post-commit hooks, reducing first-round CI failures by 60%+

Platform & Org Leadership:

• 🛡️ Led ads post-rendering workflow reliability (Newsfeed, Permalink, Reels), protecting $20–40M in revenue
• 📉 Redesigned iOS/Android metrics reporting: 30% fewer events, 10% smaller payloads at billions-per-day scale
• 🚀 Drove org-wide Swift adoption from 5% to 50% of new code; authored 10,000+ lines of new Swift in 4 months; aligned 150+ iOS engineers
• 🔓 Unblocked a cross-platform migration stalled for 1.5 years, launching it in 3 months

Snap Inc. | March 2019 - June 2025

iOS Lead — Snapchat (400M DAU, 800M MAU)

• 👥 Led the team managing key pages serving 150+ million iOS DAU
• 💰 Delivered cost-saving projects reducing media upload/retrieval/storage cost (S3/S4) by $10M+ annually
• 💾 Led rewrite of the entire backup stack onto a cross-platform TypeScript solution: 7 months dev, 11 months rollout; backup-incomplete rate 6% → 4.8%; 350M+ snaps processed daily; WWAN upload data cut 40%
• ⚡ Owned Memories playback latency across 4 billion monthly views; improved P90 image viewing latency 40% and P90 video viewing latency 60% via prefetching
• 🏗️ Led the Matcha project: rewrote 200,000 lines around proper dependency injection — 10 engineers, 30+ teams, 100+ libraries; won the company-wide Technical Excellence Award
• 📝 Built a story editor feature (0 to 1): 50M+ daily users
• 🌱 Mentored 5+ engineers to E5 promotion; led high-severity ‘code red’ P0/P1 incident responses

• 🚁 Launched Snapchat’s flying drone camera, Pixy (pairing, transfer, preview flows)
• 📷 Major contributor to Spectacles V3 launch (dual cameras); rewrote the data transfer stack (BLE, Bluetooth Classic, WiFi)
• 🏭 Owned factory QC app code & release pipelines (China/Taiwan manufacturing)

Earlier Experience

BMW Technology Corporation | May 2017 - March 2019 — Senior iOS Engineer

• Owned POI search and trip management; built a unified map-provider interface (HERE, Apple Maps) across markets
• Shipped the first-gen in-car head unit (Node.js/TypeScript) and a cross-platform after-sales tab (React Native); partnered with the ML team on personalized promotions

Citigroup | April 2016 - May 2017 — iOS Engineer, Payments Team

• Owned cards, bill payments, and account management features; built calendar import/export and Apple Watch account summaries; integrated credit score reporting; prototyped Siri-based payments

🛠️ Personal Projects

• NewsAI — AI news aggregator/summarizer built for fun: multi-agent development with Cursor, Claude API for summarization, Vercel hosting, Neon serverless DB, MailerLite subscriptions
• This site — built with multi-agent development workflows, hosted on GitHub Pages, deployed via GitHub Actions

🎓 Education

Master of Science in Electrical Engineering
University of Colorado Denver | Denver, CO

Bachelor of Engineering in Electronics Engineering
Beijing Technology and Business University | Beijing, China

🌟 Why Work With Me?

Technical Excellence

I don’t just write code – I architect systems that scale. From 4 billion monthly video views to self-improving agent loops serving 100+ engineers, I’ve proven I can handle complexity at any scale.

Leadership That Delivers

Leading 10-person teams across 30+ partner teams taught me that great code alone isn’t enough. Communication, planning, and bringing people together around a shared vision – that’s where impact happens.

Impact-Driven Mindset

Every line of code should solve a real problem. Whether it’s $5M+/yr in savings from a codemod agent or cutting latency by 60% for 150M daily users, I focus on measurable outcomes that matter to users and the business.

Continuous Learning

Tech moves fast. I stay ahead by constantly learning – this blog’s 3×/week bilingual deep-dives on LLMs and agents are proof. I bring that curiosity to every project.

📄 Download Resume

Interested in working together?

📥 Download Full Resume (PDF)

Or reach out directly:

📧 Email: sheldon.wang777@gmail.com
📱 Phone: 720.772.9666
💼 LinkedIn: linkedin.com/in/sheldonengineering

Last Updated: September 2026