Selected systems · 2023–2026

Work

Projects that turn ambiguous problems into tools, experiments, and dependable workflows.

AI evaluationAutomationFull-stack

Four selected projects

Built to answer a real question.

Each project starts with a concrete workflow and chooses technology around the people who need to use it.

01 · AI evaluation

Code Grading Evaluation Platform

A repeatable Python workflow comparing OpenAI and Gemini grading decisions with rubric expectations and human review. It surfaces inconsistent results and supports threshold-based intervention.

02 · Route intelligence

AI Perfect Walk

A route-planning proof of concept combining Google Maps services with Gemini recommendations to generate walking and running routes from natural-language preferences.

03 · Workflow automation

Jira Automation Toolkit

Python tooling for registering and triggering Jira actions through external APIs, reducing reliance on third-party automation products and making operational flows easier to control.

04 · Data systems

Inventory & Order Management

A database-backed internal application for inventory and customer order workflows, built around relational schema design and dependable Java-to-MySQL data access.

Working principles

Reliable after the demo.

01

Start with the workflow

Understand the decision, user, and failure mode before choosing the stack.

02

Keep humans in control

Make AI outputs inspectable and add review where uncertainty has consequences.

03

Measure what matters

Evaluate consistency and usefulness, not only whether an output looks convincing.

Next Experience