Build Log
My day job is confidential (marketing attribution stuff), so here's what I'm actually excited to share— the random projects I build because they seem fun, experiments that went nowhere, and side quests that turned into something cool.
experiment: Updated folder structure and DS
Refactor: Revamped folder structure for better module organization, resolved "src" naming conflicts, and solved some basic DSA problems in the terminal.
feat: Day 2 — Safe factorial (TDD), 'touch' discovery, exceptions in pytest
Day 2: Learning some new cool commands I didn’t realize you can directly use `touch` to create a file instead of `vi ` and doing `:wq` every time — should have known that! Work done: - Wrote a safe factorial function using TDD - Added tests in pytest - Practiced raising/catching exceptions and asserting them in tests Outcome: green tests, faster CLI workflow with `touch`, and better comfort with pytest exception handling. Commit: | Status: Green (Completed) | Today: Yes
experiment: Day 1 of ai-learning-journey, Environment setup, a lot of CLI
Creating the environment through Command line It's been so long since I used only command line for things like git (I mainly use GitHub Desktop) that I forgot most of the commands. Today's challenge achieved the following: 1) Create files and use git through the command line 2) Start with the basics of pytest (need to get stronger in pytests) 3) Start this journey of learning to hone my skills Outcome: repo initialized, environment ready, and my first green test with pytest.
feat: interactive HTML content manager with live preview and export
Built a comprehensive content management system to eliminate manual HTML editing for portfolio updates. Features include intelligent parsing of existing commit-style entries, real-time preview, and automated HTML generation. The tool maintains the git-aesthetic while providing a user-friendly interface for adding work entries, projects, publications, and photos. Includes backup/restore functionality and works entirely client-side for privacy and speed.
feat: auto-commit message generator using OpenAI API
Tired of writing "fix bug" for the 100th time. Built a tool that analyzes git diffs and suggests meaningful commit messages. It's surprisingly good at understanding what actually changed. Works with conventional commits format.
build: GPU price tracker with Sunday evening insights
Web scraper that tracks GPU prices across retailers. Not because I need a GPU, but curiosity about price patterns. Discovered prices drop significantly on Sunday evenings. Data visualization shows clear weekly cycles.
wip: LinkedIn buzzword blocker chrome extension
Started as a joke about "thought leadership" posts, but actually becoming useful. Blocks posts containing configurable buzzwords. UI is terrible but functionality works. Considering open-sourcing if I can make it presentable.
feat: personal website publication embedding system
Added functionality to embed research publications directly into website. Multi-strategy approach with iframe, proxy, and fallback content. Handles X-Frame-Options blocking and CORS restrictions gracefully.
fix: failed AI commit prediction experiment
Tried to build AI that predicts which commits would break production. Trained on Git history and deploy logs. 90% accurate but just predicted "probably fine" for everything. Turns out I don't break production that often. Good for confidence, bad for the model.
feat: automatic downloads folder organizer script
Python script that organizes Downloads folder by file types and creation dates. Went from digital chaos to actually findable files. Small wins count. Friends started asking for copies after seeing my clean desktop.
refactor: meal planning app in production with 20+ users
Simple meal planning tool because I was tired of deciding what to eat. Suggests meals based on what's in my fridge (manually entered, not magic). Started using it myself, shared with friends, now 20+ people use it regularly.
experiment: learning Rust vs Python performance comparison
Rewrote Python script in Rust because everyone says it's amazing. 10x faster execution but 5x longer development time. Borrow checker and I are still not friends, but starting to see why people love it.
feat: personal data pipeline for life metrics analysis
Aggregates data from fitness tracker, bank account, and calendar. Weekly reports about habits. Spend 40% more on food when working late. Most productive on Tuesdays for unknown reasons. Data-driven self-awareness.
feat: biomedical RAG system for research paper analysis
Built system that reads millions of research papers and answers scientists' questions. Spent weeks fighting PDF parsers and text extraction. Final system understood protein structures and gene interactions. Academic research is slow but oddly satisfying.
docs: published CRISPR guide prediction research paper
Co-authored paper on using ML for CRISPR guide prediction. Not my main expertise but learned computational biology. Peer review process is brutal but educational. Now have academic publication credentials.