I'm a software engineer who builds end-to-end web platforms, real-time distributed communication systems, and security-critical verification pipelines. My work centres on robust backend architecture, strict typing, database-level isolation, and pragmatic AI integration.
- Currently building multi-engine forensic verification and real-time AI collaboration tools
- I test what I ship, with full vector coverage on critical modules
Real-Time Multimodal AI Group & 1:1 Conversational Platform
- Unified conversation model: humans and streaming AI participants share the same schema primitives across private 1:1 chats and multi-user group rooms
- Synchronized multi-client streaming: a Deno Edge Function streams over SSE, flushing tokens (~400 ms cadence) into a shared Postgres row that every member tracks via Supabase Realtime
- Zero client-side key exposure: provider keys live only in edge function environments
- Fail-closed two-stage moderation (pre + post): classifier faults reject publication instead of passing through
Full-Stack Luxury E-Commerce Ecosystem with Real-Time Deal Engine
- Interactive AI Shopping Concierge plus bidirectional Socket.IO deal broadcasts
- Dynamic promotion engine with server-side percentage coupon evaluation, Zustand cart state, and automated PDF invoice generation (PDFKit)
- RBAC with JWT authentication and Bcrypt password hashing
Autonomous Multi-Engine Forensic Identity Verification & Hackathon OS
- Multi-stage forensic trust pipeline: UIDAI Verhoeff Dihedral
$D_5$ checksum validation, 128-dimensional facial landmark matching (cosine similarity), and Error Level Analysis (ELA) for tamper detection - Collision graph engine that intercepts Sybil attacks and credential reuse across aliased registrations
- Event operations module with HMAC-SHA256 digital passes, real-time ticket scanning (<150 ms verification) and Z-score normalized judging rubrics
- 69/69 test vectors passing across core forensic modules
Languages
Frontend
Backend & APIs
Databases & Cloud
AI & Forensics
Security
Testing & Tooling
| Principle | What it means in practice |
|---|---|
| Zero-Trust Security | Postgres Row-Level Security on all user-facing tables; no keys or secrets in client bundles |
| Fail-Closed Pipelines | Safety guards and classifiers reject or quarantine on upstream errors, never pass by default |
| Synchronized Real-Time State | WebSockets plus database-driven Realtime CDC / broadcasts keep every client consistent |
| Type Safety | Strict TypeScript across client state, API boundaries, and database query layers |
I'm open to conversations about full-stack systems, real-time architecture, and applied AI.
