Case Studies

Projects & Experience

A closer look at how I approach different problem spaces—from data platforms and streaming systems to IoT prototypes. Each section includes architecture highlights, metrics, and follow-up material.

NYC Taxi Incremental Processing Digital Twin for Wildfire TransNithya Mobility Smart Waste Segregation
Wildfire digital twin dashboards
  • RoleData Engineering & Analytics
  • StackPython, Postgres, Debezium, Kafka, DuckDB, Streamlit
  • Year2026

Incremental Data Processing for NYC Taxi Ride Analytics

Research question: Can Incremental Batch Processing provide meaningful advantages compared to Full Recomputation for large scale datasets?

Change Data Capture Incremental ETL Interactive Analytics
  • Capture row-level changes from Postgres using Debezium, stream to Kafka, and apply incremental loads into DuckDB.
  • Compare Incremental Batch pipelines against Full Recomputation using an experiment harness measuring Latency, Delta Scalability, and Resource Utilization.
  • Build a Streamlit dashboard to visualise ride metrics and run A/B evaluation scenarios.
  • Report trade-offs and thresholds where incremental methods outperform full recompute on NYC Taxi workloads.
Report ▶︎ Demo
Wildfire digital twin dashboards
  • RoleData platform & visualisation
  • StackPython, PostGIS, Mapbox, React
  • Year2025

Digital Twin for Wildfire

Portugal-wide situational awareness platform that fuses historical fires, meteorological feeds, and satellite telemetry to help civil protection teams plan resources during wildfire season.

Spatial ETL Predictive modelling prep Interactive dashboards
  • Normalised 20+ years of fire perimeter shapefiles and land-cover rasters into a consistent temporal data cube (2 TB).
  • Built ingestion workers that reconcile Copernicus satellite tiles with national weather feeds to refresh the twin every 15 minutes.
  • Designed Mapbox dashboards for risk overlays, control lines, and resource simulation.
Detailed Report ▶︎ walkthrough (video)
TransNithya app screens
  • RoleFull-stack engineer
  • StackSpring Boot, Angular, MongoDB, AWS
  • Year2021

TransNithya – Urban Mobility Assistant

The aim is to connect public transportation with citizens by surfacing real‑time demand signals so operators can plan capacity and run services more reliably.

Trip planner APIs Multi-tenant UI AWS deployment
  • Designed REST APIs for routes, pricing, and demand forecasting with Spring Boot + MongoDB aggregation pipelines.
  • Built Angular components for live route comparison, fare breakdowns, and saved commutes with JWT auth.
  • Containerised services with Docker and deployed on AWS EC2 with GitHub Actions for automatic image pushes.
Smart waste segregation prototype
  • RoleEmbedded & ML integration
  • StackESP32, Python, TensorFlow Lite
  • Year2021

Smart Waste Segregation

Conveyor-based prototype that separates organic, recyclable, and metallic waste streams using sensors + lightweight vision models.

IoT sensors Edge ML Actuator control
  • Implemented inductive and moisture sensors to automatically detect metal and wet waste, triggering pneumatic diverters.
  • Trained a TensorFlow Lite CNN to classify plastic vs. glass objects from a top-mounted camera at 30 FPS.
  • Built OTA diagnostics dashboard to monitor bin capacity and maintenance alerts.