01 / AI & ML
Implemented project
Compliance Assistant
Document-grounded Q&A with source citations, local embeddings, persistent retrieval, and an extractive fallback. Includes evaluation tools and a separate LoRA training workflow.
THE PROJECT WORKBENCH / PUBLIC CODE & LIVE BUILDS
Follow the
rabbit holes.
From useful AI to data pipelines, research experiments, and a website for my dad. Open the code, explore the decisions, or try a live build.
16 projects to explore
02 / DATA SYSTEMS
Engineering prototype
Real-Time Medallion Lakehouse
An e-commerce event pipeline with Kafka ingestion, Spark validation, Iceberg storage, and Polars transformations. Invalid events take a quarantine path before dbt builds analytics tables.
03 / RESEARCH
Research prototype
4D Perception Engine
LiDAR-to-camera projection, depth-to-point-cloud conversion, and heuristic hard-example scoring. NumPy/CuPy paths and an interactive dashboard make the geometry inspectable.
04 / DATA SYSTEMS
Local data platform
NYC Taxi Lakehouse
A local pipeline that cleans taxi trips, separates outliers, builds DuckDB analytics tables, and exposes demand, fare, and revenue views in a Streamlit dashboard.
05 / AI & ML
MLOps prototype
SensorOps
A maintenance demonstrator linking sample-data ingestion, experimental models, model tracking, an inference API, and optional document retrieval.
06 / RESEARCH
M.Sc. research
Reproducible Forecasting Benchmark
A controlled quantum-versus-classical forecasting study with temporal evaluation, cached features, traceable artifacts, and restartable experiment sweeps.
07 / WEB & 3D
Live website
A Research Archive, Brought to Life
My dad’s academic website: a searchable research archive, document readers, a Blender/Three.js bee experience, and a static deployment on Hostinger.
08 / DATA SYSTEMS
Platform prototype
Carbon Intelligence Platform
Emissions and air-quality ingestion, curated carbon-pricing data, dbt models, and a dashboard with simulated fallbacks. Includes Terraform definitions for a GCP analytics stack.
09 / RESEARCH
Analysis notebook
Flight Delay Analysis
A notebook exploring flight data, engineered features, and linear, random-forest, and XGBoost models. An exercise in comparing approaches and understanding their assumptions.
10 / AI & ML
Streaming ML prototype
Transaction Fraud Pipeline
A proof of concept for replaying transaction records, training an XGBoost classifier, and connecting stream scoring to an API.
11 / AI & ML
Sensor ML prototype
IoT Maintenance Pipeline
Simulated sensor events flow through Kafka and Spark, with Random Forest anomaly classification and a FastAPI prediction interface.
12 / AI & ML
ML pipeline prototype
Customer Churn Pipeline
Categorical preprocessing and logistic regression meet CSV-to-Kafka replay, Spark inference, and a FastAPI scoring endpoint.
13 / DATA SYSTEMS
Prototype · synthetic data
Stock Market Streaming Lab
Simulated stock prices move through Kafka into Spark Structured Streaming for sliding-window averages and console output. Model training and API serving are extension points.
14 / DATA SYSTEMS
Prototype · heuristic logic
Logistics Streaming Prototype
Inventory events become reorder suggestions through modular forecasting and decision functions. A Kafka/Spark experiment with a simple stock-threshold heuristic.
15 / AI & ML
Architecture scaffold
Industrial Knowledge Assistant
Simulated sensor events, Kafka, and a maintenance-question API, with extension points for document retrieval and language-model answers.
16 / AI & ML
Experimental template
Energy Forecasting Starter
An energy-modelling template with XGBoost and LSTM training, CSV event replay, and starter Kafka/Spark and FastAPI integration.
No projects match that combination. Try another tool or topic.
A workbench is always evolving. Project labels distinguish implemented work, research prototypes, and early scaffolds; repository availability does not imply a live service.
Back to the selected builds ↗