Data Science & AI · Leiden University (LIACS) · Honours Programme
Most people learning AI stop at calling the API. I want the part below it — why a retrieval pipeline returns the wrong chunk, why an agent loops, what the maths is actually doing.
Backend AI Engineering intern at flyrank.ai: RAG architecture, vector databases, chunking strategies, retrieval pipelines.
|
meme-mirror — MediaPipe · OpenCV · NumPy Webcam reads hands and face, matching meme pops up. Plain-geometry rules for poses the stock model misses, majority-vote smoothing against landmark jitter. No training. |
tech-stack-recommender — Python, zero deps Skills to ranked career paths. TF-IDF and cosine similarity written from scratch — no ML libraries — with cold-start handling. |
|
supermarket-management-system — SQL Relational schema with an IS-A staff hierarchy, weak entities, self-joins and M:N relations. Sample data and queries included. |
python-fundamentals — Python Self-contained projects: OOP class design, recursion, graph search, image processing. Fundamentals made load-bearing, not decorative. |
also: iris-classification-ai · bytebot-chatbot · all repositories →
- umaarmirzaa.github.io ·
HTML— Personal portfolio — Data Science & AI Jul 2026 - chalk ·
JavaScript— A teaching skill for Claude that fact-checks itself so it doesn't confidently teach you wrong things Jul 2026 - meme-mirror ·
Python— Webcam watches your hands and face; the matching meme pops up Jul 2026 - python-fundamentals ·
Python— Eight Python learning projects in one repo — operator precedence, loops, recursion, OOP class design, graph search, and image processing Jul 2026
core — Python · SQL · SQLite · Git
ai/ml — LLM applications · agentic systems · RAG · prompt engineering · NumPy · supervised & unsupervised learning · MediaPipe · OpenCV
foundations — algorithms & data structures · OOP · relational databases · image processing · graph theory · calculus I–III · linear algebra
Everything here is built rather than described. That's deliberate.