Alexander Wang

Education

• Relevant Coursework: Algorithms, Object-Oriented Programming, Discrete Mathematics, Databases, Machine Learning, Digital Electronics, Computer Graphics

Experience

• Building an internal search portal that lets teams across the organization discover and access enterprise data sources
• Catalogued insurance data into the internal Data Marketplace, writing SQL to profile, validate, and document source tables
• Built Power BI dashboards across multiple business data sources to support reporting and data-driven decisions

• Built and operate an AI-assisted writing platform for college applications: 200+ users and paying customers admitted to Yale and UC Berkeley (Next.js, React, TypeScript, PostgreSQL)
• Own the full stack — schema design, Stripe billing, and an AI orchestration layer with an ordered provider fallback chain over Gemini (Vertex AI and AI Studio), streaming responses, fail-closed per-feature daily quotas, and per-run token and latency logging
• Shipped a separate internal admin app with TOTP 2FA, three least-privilege Postgres roles, and an append-only audit log for support and billing operations

• Placed 2nd in a team-based mock trading competition and presented a long/short equity pitch built on fundamental analysis and market signals

Projects

• Built a C++20 engine to parse and validate MAVLink v2 drone telemetry over UDP, reaching approximately 4M frames/s with one processing thread in recorded loopback benchmarks on WSL2
• Published live vehicle state to a Python autonomy layer via seqlock shared-memory IPC, with stale-data guards for geofence and low-battery decisions; validated loss accounting and corruption rejection against a seeded proxy that injects packet drops, delays, reordering, and corruption
• Implemented a preallocated ring buffer and async-signal-safe crash handler that produced complete black-box dumps in 20/20 recorded crash tests; built a Python tool for post-mortem decoding and CRC verification

• Built a "semantic Ctrl+F" Chrome extension (Manifest V3) and web demos that match meaning rather than characters — fully on-device with no backend, using 384-dim MiniLM embeddings
• Fused literal, keyword (MiniSearch), and semantic rankings with weighted Reciprocal Rank Fusion; ran inference in Web Workers with zero-copy ArrayBuffer transfer and IndexedDB caching to keep typing jank-free

Skills & Interests