mbongowo/Data-science-Portfolio

24 reproducible spatial & big-data projects (STAC/Sentinel-2, GeoAI, Kafka/Spark, dbt) with ~800 CI-tested cores and 4 live Streamlit apps. Cameroon-focused. # Data Science Portfolio Data science work by **Joseph Mbuh**: spatial / remote-sensing pipelines and large-scale (big data) analytics, built so the same code runs the same way on someone else's machine as it does on mine. Each project lives in its own folder with its own environment, tests, CI, and documentation. The work splits into a **spatial** track (remote sensing, geospatial pipelines, GeoAI), a **big data** track (streaming, distributed processing, the modern data stack), and a set of **technique-replication** projects that rebuild the idea behind a well-known open-source repo on a new dataset or region. Every folder stands alone; pick by the role you are reading for. Across all of them the pure-numerical core runs and is tested in CI, while the heavy parts (GPU, Earth Engine, Docker/Kafka, Terraform/Azure, deploys) are documented to run on your own machine. ## Spatial track Six projects, ordered so the skill areas alternate and the later ones reuse code from the earlier ones. The deployed app imports the flagship pipeline as a dependency, and the change-detection project builds its data cube the same way the flagship does, so the projects visibly compose rather than repeat. | # | Folder | What it does | Main skill | |---|--------|--------------|------------| | 1 | `eo-monitor` | Pulls Sentinel-2 from a STAC catalogue over an area of interest, computes vegetation and moisture indices, scores anomalies against a baseline, and writes cloud-optimised GeoTIFFs. One command, no manual downloads. | Cloud-native EO pipeline (STAC, Dask, COGs) | | 2 | `access-to-care` | Travel time from each populated place to the nearest health facility over a road network, weighted by population to show who is far from care. Built around Cameroon. | Vector and network analysis | | 3 | `spatial-hotspots` | Exploratory spatial data analysis: spatial weights, global and local autocorrelation, cluster and outlier maps, with the interpretation written out and the limits stated. | …