Projects
Selected Works
AI products, full-stack platforms and a few experiments — what each one is, and the technical problem I solved in it.
AI Catalyst
An AI-assisted founder validation product that takes an idea through structured research and decision-making. I designed the system around MCP, structured outputs and clear AI/human boundaries, then took it through full-stack development and production.
Rally
An investor intelligence and matching system built around actual investment behaviour. I designed the enrichment and matching pipeline, combining deterministic filters, AI-assisted ranking and evidence-backed recommendations.
DataCopilot
A data-cleaning copilot that turns natural-language intent into deterministic transformations. I separated AI planning from execution so users could preview and verify changes before they touched the data.
Koala Be Where
My first end-to-end AI product, using computer vision to locate koalas in citizen-science photos. I worked on model training, hard cases and integrating the model into a system that had to handle messy real-world images.
SecuFlow
I turned an academic coordination model with no reference implementation into a working repository security system, then improved its performance enough to support interactive analysis.
REAMP
A multi-role workflow for real-estate media production. I used it to go deeper into full-stack system design, including shared data, role-based permissions and workflows across different types of users.
IoT Watch
A real-time monitoring system for IoT devices. I worked with event-driven data, state changes and alert logic, learning to design around continuous system behaviour rather than request-response flows.
Rhombus AI
An AI-assisted data operations product for spreadsheet transformation. I worked on making model-generated actions safer in production through validation, recovery and more deterministic execution.