What I build with

Python is the language every AI system I build runs on, from FastAPI services to PyTorch models and OpenCV pipelines.

4 projects

PyTorch is what I train models in, including the transformer segmentation models behind my IEEE Access paper.

2 projects

LangChain for retrieval pipelines and agents where its abstractions earn their keep, and plain Python where they do not.

1 project

OpenAI models behind a provider abstraction, with structured outputs and an evaluation set so the feature stays correct.

4 projects

OpenCV for real-time video pipelines on ordinary cameras: capture, preprocessing, tracking and overlay.

2 projects

YOLOv8 for detection and segmentation in production vision systems and in my published research.

2 projects

Pinecone and pgvector for retrieval-augmented generation, chosen by scale rather than habit.

1 project

n8n workflows that connect your tools and AI models, with retries, alerts and logs so failures are visible.

2 projects

FastAPI is my default Python backend for AI products: async endpoints, typed request models, and generated docs.

2 projects

Flask powers several of my computer vision and internal tools where a small, explicit Python service is the right size.

4 projects

Django with Django REST Framework for products that need auth, admin and a relational data model on day one.

1 project

Node.js backends for JavaScript-first teams and products that do not need Python's ML ecosystem.

1 project

WebSockets for live progress, streaming model output and multiplayer or bot interactions.

2 projects

.NET Core, Entity Framework and Angular on an enterprise, WCAG-compliant state platform.

1 project

Stripe on the web, in-app purchases on mobile, and the webhook handling that makes billing reliable.

2 projects

Next.js for the web products my AI systems live in, including this site, Shapes.inc, Testorim and Alif Ujala.

1 project

React across six shipped products, from AI dashboards to a state education platform.

6 projects

TypeScript on every JavaScript project I ship, front end and Node.js back end.

5 projects

Flutter for iOS and Android apps sharing one backend with the web product, including in-app purchases and store release.

1 project

MongoDB where documents fit the data: flexible AI configuration, conversation history, content.

1 project

PostgreSQL as the default database, with pgvector for retrieval when a separate vector store is not justified.

2 projects

Containers for every service I deploy, and Kubernetes when there are enough of them to need it.

1 project

AWS for backends and model serving, with the bill measured before it is optimised.

Vercel for Next.js front ends: preview deployments per branch, edge caching and analytics with no servers to run.

Playwright for end-to-end tests and browser automation, including the engine behind Testorim.

2 projects