Hello! I'm

Antonio West

Full-stack and AI systems engineer. Building production web applications, offline-first client architecture, and deterministic LLM pipelines across Next.js, TypeScript, and PostgreSQL.

Explore WorkGitHub
Full-Stack & AI

Production Systems

Deterministic web apps, offline client storage, and guarded LLM pipelines.

TypeScript & Next.jsPostgreSQL
TypeScript
PostgreSQL
Next.js 16
Offline-First

Client Systems & Architecture

Be Act ChangeProduction PWA

Document Parsing & Personalised Coaching Pipeline

Automated text extraction from multi-page PDF & DOCX uploads using pdfjs-dist and mammoth to eliminate manual data entry. Built a 23-step dynamic onboarding state machine with localStorage session recovery, strict JSON schema validation, and automated GPT-4o output regeneration.

Next.jsTypeScriptGPT-4opdfjs-distmammothshadcn/ui
Alive IndustriesChatSDK Infrastructure

Meeting Audio Summarisation & Intelligence Engine

End-to-end meeting intelligence pipeline for ChatSDK ingesting Whereby recordings through a multi-provider LLM summarisation layer. Built background workers with intelligent transcript caching and polling, persisting structured action items directly to PostgreSQL via Drizzle ORM.

TypeScriptNext.jsPostgreSQLDrizzle ORMWhereby APIshadcn/ui
Featured Product

Veribite

Founder & Lead Engineer/Bethnal Green Ventures Finalist

Sub-Second Allergen Detection for Complex Grocery Labels

Built to remove the risk of human fatigue when checking packaging labels in busy supermarkets. Shoppers scan a barcode to get an immediate, deterministic safety verdict against their specific allergens from a local cache of 70,000+ UK products, avoiding manual label audits with zero network latency.

70k+
UK Products
<1s
Offline Verdict
14 EU
Allergen Rules
Next.jsTypeScriptOffline-FirstBarcode API14 EU AllergensTailwind CSS
Daily Feed

Personalised Dashboard

Weekly shopping summary, quick barcode scanner, and live allergen scan history.

Personalised Dashboard
Instant Catalog Search
Verified Safe Verdict
Cross-Contamination Alert
Definitive Allergen Match
Swipe or drag to preview (1 of 5)

Systems & Benchmarks

Applied Machine LearningR² 0.99 | RMSE 0.1577

Molecular Toxicity & Property ML Pipeline

High-throughput machine learning pipeline predicting molecular properties and compound toxicity with TreeSHAP explainability and partitioned Parquet data streaming via PyArrow for memory efficiency.

PythonXGBoostTreeSHAPPyArrowParquetScikit-Learn
Graphics & WebGL Systems-75% VRAM Reduction

WebGL Shader & Texture Memory Optimizer

Client-side utility that packs discrete PBR material channels (Roughness, Metallic, AO, Height) into single 32-bit RGBA textures, cutting browser VRAM footprint by up to 75% while locking rendering at 60 FPS.

WebGL 2.0GLSL ShadersTypeScriptThree.jsPBR Packing