Huzaifa Nasir.

Algorithms, systems and machine learning, alongside the full-stack work that turned them into things people could actually use.

Shipped AI-backed web features and the automation that connected them to the rest of the business.

• Built model-backed features in Next.js 15 and TypeScript, wiring inference endpoints into typed server actions and streaming responses to the client.
• Automated cross-department workflows in n8n, connecting services that until then required someone to copy data between them by hand.
• Set up GitHub Actions pipelines running type checks, linting and preview deployments on every pull request.

Building the retrieval and agent systems behind OneScreen products, from architecture through to production.

• Design retrieval-augmented chatbots across the full pipeline: document chunking, embedding strategy, vector search and reranking, with answers traced back to their source passages.
• Build multi-agent workflows in LangGraph where specialised agents pass state through a shared graph, with retry and fallback paths so one failing tool call does not collapse the run.
• Fine-tune and optimise transformer models in PyTorch, trading model size against latency to keep inference inside the product response budget.
• Orchestrate data synchronisation between internal services in n8n, replacing manual handoffs with scheduled pipelines that surface their own failures.
• Take features from prototype to production: evaluation harnesses, failure-mode testing, and the deployment path they ship on.