Brodin OS is a self-built, full-stack platform designed as a single "operating system" for my life as a university student. A FastAPI + SQLAlchemy modular monolith serving a React/TypeScript/Tailwind SPA with lazily-loaded apps behind a shared sidebar shell. Its flagship Academic app syncs my Brightspace iCal feeds, deduplicates deadlines with fuzzy matching, parses syllabi with an LLM, and auto-generates T-10/T-5/T-2 study milestones, presented through agenda, kanban, and grade-tracker views. Every app shares one AI backbone (OpenRouter via the OpenAI SDK) and one design philosophy: AI features degrade gracefully to manual workflows instead of breaking. Alongside Academic, three more apps complete the toolkit:
I log meals by snapping a photo: an AI vision model estimates calories and protein from the image (with my own text context treated as authoritative), with manual entry always available as a fallback. Daily summaries, calorie/protein goals, weight tracking, and a history strip make it a lightweight but complete personal nutrition diary.
It scrapes real Food Basics flyer deals through the store's public JSON API, then an LLM generates a 7-day, 2-meals-a-day plan sized to my calorie/protein targets and priced to stay under my weekly budget. A second LLM pass consolidates all ingredients into whole purchasable products ("1 dozen eggs", not "2 cups cooked rice"); a real, cart-ready grocery list that reuses leftovers and prioritizes on-sale items.
Uploaded audio is preprocessed directly in the browser: unsupported formats (FLAC, WMA, or oversized files) are decoded via the Web Audio API and converted to MP3 using lamejs. The audio is then chunked, transcribed by a model, and stitched together in memory with guards against repetition loops. Finally, a cleanup pass fixes punctuation, capitalization, and filler words, while keeping both raw and polished drafts editable side-by-side.
Co-founded an ongoing initiative focused on engineering compact, modular IT infrastructure and mounting hardware tailored for home laboratories and network setups. Modeled parametric components and structural assemblies in SolidWorks, while developing photo-realistic product renders in 3ds Max and Arnold to guide brand identity and early interest. Currently fabricating physical prototypes, iterating on mechanical tolerances, and preparing the ecosystem for initial production runs.
Re-engineered a stock Creality Ender 6 into a high-performance Voron CoreXY hybrid system by overhauling its mechanical kinematics, power distribution, and embedded control stack. Integrated modern 32-bit microcontroller hardware, rewiring sensor signal paths and stepper motor drive buses with optical endstops and high-current drivers to eliminate noise and resonance. Compiled and calibrated custom real-time Klipper firmware, implementing digital signal filtering via input shaping resonance testing and pressure advance to optimize closed-loop velocity profiling.
Architecting an end-to-end full-stack web application designed to collect, process, and analyze real-time automotive market datasets. Built asynchronous Python data ingestion pipelines with robust scraping routines that parse, validate, and normalize unstructured listing data into a structured schema. Currently developing dynamic front-end query interfaces and integrating LLM inference endpoints via REST APIs to perform automated data summarization, anomaly detection, and valuation analysis on ingested records.