Staff-level AI and backend infrastructure engineer with 6 years of experience and over $10M in business impact, working across software and hardware engineering, ML, security, product, and infrastructure.
People leadership: Grew and led a team of 7. Enjoy working with sharp, low-ego engineers.
Interviewing: Have run dozens of algorithm and system-design interviews.
Track record: Consistently ship reliable systems at scale through cross-team work.
Technical Expertise
Languages & Frameworks: Golang, Python, gRPC, GraphQL
Cloud & Infrastructure: GCP, AWS, Kubernetes, Vertex AI
AI / ML: vLLM, Claude, Pinecone, Arize
Data & Observability: MongoDB, Grafana
Work Experience
Samaya AI, London
- Built the knowledge ingestion pipeline, chunking and embedding strategy, eval harness, RAG, and the AI infrastructure behind them
- Took ML ideas from 0 to 1 into shipped features: query decomposition, deep research, and earnings reports
- Cut time-to-first-token from 100s to 10s (90% faster) and scaled the platform from 10 to 10,000 daily active users
- Onboarded customers including Morgan Stanley, BlackRock, Citi, and Point72, and helped set pricing
- Built and ran the demo for our $43.5M Series A
Tesla, Shanghai & Texas Gigafactories
- Owned high-voltage systems across Model 3/Y/S/X, Powerwall, Megapack, and Supercharger
- Iterated chip designs for Cybertruck and Megapack on a tight 2.5-month deadline
- Worked closely with vendors worldwide (CATL, BYD, TI, COSMX, HYC, Pegatron, Jabil), traveling on-site for design reviews, inspections, troubleshooting, and bringing down defect rates
- Trained manufacturing, process, and quality teams to run their own debugging and fault analysis; got the China battery line running 3x faster than the US line
- Delivered the Cybertruck battery charger 4 weeks ahead of schedule
- Brought up 50 new manufacturing stations, maintained 200+, and managed a $500k annual budget
Atlassian, Mountain View, CA
- Built the auth sidecar securing all traffic between Atlassian microservices and staff
- Rewrote internal tools (LDAP cache, rate limiter, deduper) from Java, TypeScript, and Python into Go
- Ran on-call for a 99.99%-availability internal PaaS hosting 1,400+ Atlassian services
- Built a service generator that scaffolds new services with Kubernetes, SOX compliance, and CI/CD out of the box
Tools: Splunk, SignalFx, Jira, Confluence, AWS S3, DynamoDB, Lambda, Postgres, Spinnaker
Education
Johns Hopkins University
Coursework: Cloud Security, Software Vulnerability Analysis, Computer Forensics, Network Security, Privacy Laws, Risk Management, Ethical Hacking, Modern Cryptography
University of Rochester
Joseph C. Wilson “Change” Scholarship, Dean’s List
Coursework: Signal Processing, Robot Control, Semiconductors, Microprocessors, C++, Java, Assembly, VHDL, MATLAB
Minor: Mathematics and Economics
Projects & Research
Research
Behavioral Biometrics Security Research, Unknot.id
- Built adversarial attacks against Android behavioral-biometric authentication
- Showed a shadow-model attack that broke the system in under 1,000 queries
- Hardened the target LSTM models against those attacks with CleverHans
Cognitive Security Modeling, Johns Hopkins University
- Modeled human security decision-making in Python using Instance-Based Learning Theory and ACT-R memory architectures
- Processed and analyzed 40GB+ of behavioral data to surface security-relevant patterns
- Worked on an interdisciplinary team bridging computer science and behavioral psychology
Technical Projects
Ontra.ai
- Replaced a thin GPT-wrapper (Ruby on Rails) with a real ML service doing retrieval, cross-encoding, and prompt evaluation
- Saved lawyers thousands of human-in-the-loop hours previously lost to poor model output
Agemo Codewords
- Full-stack web app for automated website building, built with Node.js and shadcn/ui
Rippling Global Expansion Framework
- Built the global-expansion framework for Rippling’s payroll product with Django, MongoDB, and React
LeetCode
- Solved 1,000+ problems and raised my contest rating from 1460 to 2109 in seven months
Military Drone System, U.S. Army Contract
- Designed and built a drone-based supply-drop system for Army training exercises
- Used 3D CAD, additive manufacturing, and Arduino-based controls
- Fielded at the National Training Center and Joint Readiness Training Center
Battle Map Gaming Platform
- Real-time multiplayer battle-map platform over WebSockets, with live state sync across players
- Full-stack and cloud-hosted on Google App Engine
Security Research
- Docker privilege escalation: used the Dirty COW race condition for container-to-host escape
- IoT exploitation: network analysis and attacks on DJI and Bebop drones and Amazon Echo
- Malicious URL detection: ML models on the PhishTank and DMOZ datasets, evaluated with ROC curves
- Software vulnerability analysis: threat modeling and reverse engineering with objdump, ldd, nm, and strace
- CryptoDoneRight: contributed server-hardening tools to cryptodoneright.org
Interests
Gaming: built a gaming platform; top 0.3% League of Legends player
Music: guitarist with live concert experience; university carillon society
Athletics: stand-up paddleboarding, amateur boxing, 10+ pull-ups
Hardware: headphone amplifier, windshield-wiper controller, custom PC builds
Links
- Website: zuo.vercel.app
- LinkedIn: linkedin.com/in/mrmightymango
- GitHub: github.com/MrMegaMango
- LeetCode: leetcode.com/mangoman
- Language Exchange: echinglish.weebly.com
- Research Profile: behavior.isi.jhu.edu