Skip to content
The stack

Technologies

What I reach for and how deep I go. Each links to a short note and the projects where I've used it.

Languages

Java4 projects
8–21 — multi-agent simulation engines and backend services.
Julia1 project
High-performance numerical and agent-based simulation.
Python10 projects
Primary language — GenAI pipelines, data work, services, and optimisation.
SQL
PostgreSQL/PostGIS, Oracle, BigQuery — schema design and query tuning.

Frameworks & libraries

Agent Development Kit (ADK)1 project
Vertex AI agent framework — tool-using agents with deterministic tools.
Celery1 project
Distributed task queues with RabbitMQ for async job delegation.
FastAPI
Async Python web framework with type-hint-driven validation and OpenAPI docs.
Flask
Minimal, unopinionated Python web framework for APIs and services.
Google Gemini2 projects
Gemini models via the Google GenAI SDK — classification, extraction, and agents.
LangChain
Composable chains, agents, and provider integrations for LLM applications.
MASON2 projects
Java agent-based modelling toolkit — battlefield and epidemiological simulation.
pandas
Data wrangling and analysis across ML and reporting work.
Polars1 project
Fast DataFrame processing for large tabular pipelines.
Spring Boot1 project
REST services and persistence with Hibernate over Oracle.

Data stores

BigQuery4 projects
Warehouse-scale analytics and feature data for ML pipelines.
PostgreSQL1 project
Primary relational store, with PostGIS for geospatial scoring models.

Infrastructure & cloud

AWS1 project
EC2-hosted simulation workloads.
Google Cloud Platform4 projects
Vertex AI, BigQuery, GCS, Cloud Composer, Workbench — plus org admin and cost governance.
Vertex AI2 projects
Managed pipelines, training, and model serving for production ML and GenAI.

Tooling

Cloud Composer / Airflow1 project
DAGs orchestrating production ML pipelines.
Docker3 projects
Reproducible local envs and deployable build artifacts.
Langfuse
Tracing, evaluation, and prompt management for LLM applications.
MLflow
Experiment tracking, model registry, and reproducible packaging for ML runs.

Practices

MLOps2 projects
Model registry, CI/CD, deployment, monitoring, and self-service tooling for data scientists.
Optimisation & OR4 projects
Integer programming (MIP), genetic algorithms, and solver-based scheduling.
Prompt engineering2 projects
Iterative prompt tuning with domain experts, plus automated re-scoring and evaluation.
Time-series forecasting2 projects
ARIMA/ARIMAX and AutoML forecasting; vendor-model benchmarking.