Hello, I'm Alvin
Full-Stack Developer & AI Developer
Building scalable web applications and intelligent AI-driven solutions to power the next generation of digital experiences.
Alvin Xu
AI/ML & Full-stack — M.S. in Computer Science
- AI assistant/character demo: streaming dialogue with long-term memory; local 7B for dev, API as fallback
- Document Q&A you can measure: embeddings in Postgres (pgvector), combine semantic + keyword search, small offline test set
- Auth & permissions: Keycloak OIDC integrated with NestJS; role checks (RBAC) and basic audit logs
- Delivery & reliability: Docker + GitHub Actions; structured logs + light tracing; background jobs with retries/backoff
- Backend-first delivery: NestJS/Node services · PostgreSQL/Prisma modeling
- Practical AI integration: OpenAI + local 7B calls · retrieval pipelines
- Frontend that ships: Next.js/React · Tailwind/shadcn for clean UIs
- Deploy & debug: Docker/Nginx · AWS/Azure basics · logs/traces for triage
- Ship small; verify with metrics & evals (latency, cost, quality)
- Keep designs simple & explicit: clear APIs, tests, and documented trade-offs
- Assume failure; make it safe: idempotency, timeouts, retries/backoff, rate limits
- Privacy & security by default: least privilege, secrets hygiene, audit trails
Full-stack investment platform with SwiftUI iOS app and Node.js/Express backend, integrating Finnhub/Polygon APIs for live market data. Supports watchlists, stock search, interactive charts, and P/L tracking.
Modeled DB2 schema for multi-station hydrology time series; designed indexes for range queries; built KNN/Naive Bayes/SVM baselines for forecasting.
Scalable perceptual-hash indexing with tolerant matching for deduplication and archive retrieval across large media datasets.
Built a file system module and disk driver in C: on-disk FAT structures, partitioning/mounting, and a high-level API for read/write operations.
Implemented NER with BiLSTM/Transformer and GloVe embeddings; improved accuracy and enabled large-scale text analytics & tagging.