Chemical engineer turned AI engineer, and a generalist by trade: pipelines and warehouses, computer vision and time series, retrieval systems, and the dashboards that put the result in front of someone. Two degrees — BSc Chemical Engineering (Universidad de Los Andes, Bogotá) and MSc Artificial Intelligence (UPF Barcelona + Univerza v Ljubljani, 8.54/10) — plus a first-author paper under review at Computers & Geosciences.
garasnote.github.io · CV (PDF) · LinkedIn · jp.osorio4102@gmail.com
What I work across
- Research / computer vision — snowy: snow cover from 12 fixed webcam stations, 0.90 mIoU and +0.10 AUROC over the baseline; paper under review at Computers & Geosciences, dataset on Zenodo. Repro-INPFormer: reproduction and extension of a CVPR 2025 anomaly-detection method.
- Data engineering — hiresignal: 231,859 postings from eleven applicant-tracking systems, deduplicated on keys that are not unique, modelled into a star schema with dbt, tests and lint wired into CI.
- Big data / mobility — modes-of-transportation-ds: transport-mode inference from cellular network data (FRI Data Science competition 2025); big-data: end-to-end analysis of NYC TLC trip records.
- Retrieval / LLM — rag-service: BM25 and a dense retriever over the job corpus, answers with citations, every configuration benchmarked on labelled questions — the table keeps the row where the fancier one lost.
- Chemical engineering + data — SIVIGILAMeta: SIVIGILA and RIPS public-health data for chemical-security questions, which is where both degrees meet.
- Desktop tooling — dms-workspace-manager: a niri / Dux Shell plugin in TypeScript, in daily use.
Toolkit — Python · R · SQL · MATLAB · Bash — pandas, NumPy, GeoPandas, scikit-learn, PyTorch, Matplotlib, Seaborn, Plotly — DuckDB, dbt, Parquet, Docker, Git, pytest, GitHub Actions — Power BI (measures and DAX), Streamlit, Excel + Apps Script.