Director of Software · Technical Architect · AI Researcher Ireland 🇮🇪
I build software systems, lead engineering organisations, and occasionally disappear down research rabbit holes.
My work sits at the intersection of software architecture, technical leadership, developer productivity, and applied AI. I enjoy solving difficult problems, building high-performing teams, and exploring how emerging technologies can be applied in practical environments.
Research investigating how fine-tuning small language models for tool use impacts their broader capabilities in edge environments.
Key areas of investigation:
- Tool-use performance and function calling
- General reasoning capability
- Edge deployment constraints
- Quantisation effects on model behaviour
- Benchmark design for agentic systems
- Evaluation of agentic behaviour on resource-constrained hardware
Technologies: Qwen, GGUF, llama.cpp, BFCL, Python
Fine-tuned Qwen 3.5 4B model specialised for API generation and tool-use tasks.
Highlights:
- Trained on synthetic tool-use traces
- Released in GGUF format for efficient edge deployment
- Evaluated across reasoning and tool-use benchmarks
- Part of ongoing research into agentic behaviour in small language models
Model:
🤗 https://huggingface.co/CFitzsimons/Qwen3.5-4B-APIGen-MT-5k-F16-GGUF
Led the design and delivery of a highly scalable custom platform for a Fortune 500 organisation.
Responsibilities included:
- Technical leadership across ~30 engineers
- System architecture and technical strategy
- Delivery planning and execution
- Cross-team engineering alignment
Focus areas included scalability, reliability, observability, and long-term maintainability.
Exploring how language models can be adapted and deployed in constrained environments without sacrificing practical capability.
Areas of focus include:
- Fine-tuning small language models (SLMs)
- Edge deployment strategies
- Quantisation and model compression
- Function calling and tool use
- Agentic systems
- Evaluation and benchmarking methodologies
Understanding how developers interact with AI tools and how those interactions influence productivity, confidence, learning, and decision-making.
Investigating whether software defects can be classified through a shared framework similar to vulnerability scoring systems such as CVSS.
Exploring ambiguity in software engineering, requirements interpretation, and how experience influences technical judgement.
Presented at DEIMS, Japan.
- JavaScript
- TypeScript
- Python
- C#
- React
- Node.js
- PostgreSQL
- MySQL
- AWS
- Docker
- CI/CD
- Serverless Architectures
Throughout my career I've worked across startups, scale-ups, and enterprise environments, helping organisations navigate growth, technical complexity, and delivery challenges.
Areas where I spend most of my time:
- Technical Leadership
- Software Architecture
- Distributed Systems
- Platform Engineering
- AI Strategy & Adoption
- Engineering Effectiveness
- Team Development
Outside of work you'll usually find me:
- Running with my two dogs
- Kitesurfing (Just starting out - talk to me about it!)
- Playing Dungeons & Dragons
- Reading fantasy and science fiction
- Building things that seemed sensible at the time
- LinkedIn: https://www.linkedin.com/in/cfitzsimons
- Hugging Face: https://huggingface.co/CFitzsimons
"The most interesting problems usually begin with someone saying: 'This probably won't work.'"