You need technical knowledge to analyze scientific data but many LLM's struggle to ground their analysis in observations (raw data).

Making a dataset and training Qwen2.5vl to analyze electrochemical data.

Like today you dont need to be a programmer to use a computer, in the future you might not need to be a scientist to understand the data

Built and deployed a self-hosted infrastructure running multiple containerized services accessible from anywhere on the internet through secure reverse proxy configuration.

The system hosts media streaming, game servers, and websites using Docker containers, SSH remote management, and CloudFlare DNS routing - all running headless on Ubuntu Server from my home network.

This eliminated monthly subscription costs while providing hands-on experience with enterprise server administration, networking security, and containerized service deployment.

Designing and building a fully autonomous testing pipeline that integrates lab hardware and computer vision operating without human intervention.

The system moves physical samples, scans data codes, performs tests, and formats data for ML workflows at 15 samples per hour through coordinated Python scripts, C++ hardware interfaces, and custom communication protocols.

This eliminated tedious manual work while delivering 2-3x faster throughput and higher-quality datasets for ongoing lab research.

WIP

WIP

WIP