I work on technically demanding AI and software problems where learning systems, simulation, automation, APIs, and scientific computing meet.

My interests are less about isolated models and more about complete research systems: defining the problem, building the environment, implementing the learning or optimisation loop, instrumenting experiments, and turning the result into software that can actually be run and inspected.

Current areas of interest include reinforcement learning, multi-agent systems, computer vision, LLM/RAG pipelines, optimisation, autonomous agents, simulation, and AI-backed backend systems.

Multi-agent reinforcement-learning research using a swarm of simulated UAVs to estimate the spatial extent of an oil spill. The project combines a custom simulation environment, continuous-control DDPG, PyTorch actor-critic models, replay-based learning, and distributed sensing.

Research themes: multi-agent RL · autonomous UAVs · continuous control · environmental sensing · simulation

Automation system for producing high-resolution astronomical sky maps by controlling Stellarium programmatically. Supports location-driven generation, seasonal and temporal configuration, batch workflows, remote-control integration, and print-oriented rendering.

Engineering themes: scientific automation · external-system control · astronomy · batch pipelines · Python

Compact retrieval-augmented generation backend that connects document ingestion, vector search, retrieval, LLM response generation, and an HTTP API.

Engineering themes: RAG · vector retrieval · LLM systems · FastAPI · backend architecture

Modular asynchronous Python package for image discovery and download, with filtering, CLI and library interfaces, configuration management, structured logging, tests, and reusable components.

Engineering themes: asynchronous Python · automation · CLI design · package architecture · testing

AI / ML
├── Reinforcement Learning
├── Multi-Agent Systems
├── Deep Learning
├── Computer Vision
├── LLM / RAG Systems
└── Optimisation

Engineering
├── Python
├── PyTorch
├── FastAPI
├── REST APIs
├── Async I/O
├── Automation
└── Scientific Computing

Research Systems
├── Simulation Environments
├── Experiment Pipelines
├── Data Processing
├── Model Evaluation
└── Reproducible Workflows

I prefer projects where the interesting part is not only the algorithm.

A typical problem for me looks more like:

problem formulation
        ↓
simulation / data pipeline
        ↓
model or optimisation method
        ↓
experiment + instrumentation
        ↓
API / automation / usable system

That usually means working across several layers at once: modelling, numerical code, system integration, debugging, experimentation, and software structure.

• autonomous and multi-agent learning systems;
• local and hosted LLM inference;
• retrieval and agentic AI architectures;
• computer-vision pipelines;
• simulation-driven research;
• optimisation and computational intelligence;
• scientific automation;
• AI systems that interact with real software infrastructure.