Projects

Agentic planning + tool execution with RAG-backed retrieval, served behind intelligent APIs.

Full-stack AI application using the OpenAI SDK and modern agent-based architectures, combining agentic workflows, RAG, and vector search to create an intelligent assistant capable of handling complex, multi-step tasks. Leverages AI agents with planning and tool execution, integrates a vector database for contextual retrieval, and exposes intelligent APIs through a backend service with a responsive frontend interface.

Codebase-aware reasoning that plans, edits, and debugs across real repositories.

Claude-style AI coding agent using the OpenAI SDK and MCP servers. The agent supports multi-step planning, tool execution, and codebase-aware reasoning to assist with code generation, refactoring, and debugging in real-world repositories.

High-accuracy structured extraction from insurance documents on a cloud-native pipeline.

Cloud-native document processing platform using Python Flask microservices and Gemini AI / Vision API for high-accuracy data extraction from insurance policy documents.

Automated AWS→GCP migrations (EKS→GKE) with Terraform, improving cost and security.

Automated large-scale AWS to GCP migrations, including EKS clusters to GKE, using Terraform and custom scripts — improving performance, security, and cost efficiency.

One framework to publish and manage multi-tier SaaS listings on GCP Marketplace.

Automation framework for managing and deploying SaaS listings on the Google Cloud Marketplace across multiple product tiers.

Physically-modelled guitar tone and a scale-accurate fretboard, in vanilla JS with no dependencies.

Interactive guitar fretboard for learning the neck in standard tuning. Highlights scales by interval role, maps open chord shapes with fingerings, and drills note recall with a scored quiz. Every note is synthesised in the browser with a Karplus-Strong plucked-string model — distinct electric and acoustic voices, zero audio samples to download.