Capabilities
My work falls into three areas: agentic AI that acts inside your systems, AI models trained for a specific job, and the automation that moves data and work between them. Consulting, deployment and ongoing support run across all three. Each area has its own page; each capability within it has a page of its own.
Agentic AI
AI that does things: agents with access to your knowledge and tools, connected through MCP and put in front of your team through chat.
Agents · RAG · MCP · Chat ↓ 02AI Model Development
Models trained or fine-tuned for one job: speech recognition, LLMs, classifiers and other small models, with the datasets and evals behind them.
ASR · LLMs · Classifiers · Data · Evals ↓ 03Automation & Workflows
Workflows and pipelines built for your process, with AI steps where they earn their place, and the dashboards and alerts that show they're working.
Workflows · Pipelines · Reporting ↓Agentic AI
Moving from AI that writes to AI that acts: agents that monitor, process, route and act inside your existing stack, with people in the loop where it matters.
Agentic AI and its moving parts →MCP
MCP Integrations
Connecting agents to existing MCP servers for CRMs, ticketing and ERP, and wrapping APIs where no server exists.
Private MCP Development
Custom MCP servers for your databases, internal APIs and file stores, with authentication, scoped permissions and audit logging.
AI Model Development
General models are trained to be decent for everyone. When your work depends on words, labels or decisions a general model gets wrong, a model trained for the job does better, and often runs smaller, cheaper and on your own hardware.
AI model development overview →ASR Fine-Tuning
Speech recognition fine-tuned for your vocabulary, accents and speakers who switch languages mid-sentence, then deployed offline, on a public API or as a private internal API. I train and publish my own open-source ASR fine-tunes, and bring the same process to professional use.
See ASR fine-tuning →- Niche vocabulary: legal, medical, technical, brand and project names
- Under-represented accents
- Code-switching between languages
- Base model chosen by workload and target device
- Real or synthetic speech training data
- Retraining as your vocabulary changes
Fine-tuning
LLM Fine-Tuning
Open-weight LLMs fine-tuned with LoRA for your domain, format and tool calls, and distilled into models you can host.
Classifier Training
Text classifiers for routing, triage, intent and moderation against your own labels, cheap enough to run on every record.
Small & Specialised Models
Embeddings, rerankers, extraction models and small language models for on-premises and on-device use.
Datasets
Dataset Preparation & Blending
Training data, dataset structure, and blends of public and private data for one objective.
Training Recipes
Drafting and reviewing training plans: data weighting, epochs, settings and methodology.
Model Versioning & Releases
Version control, backup and a release cadence for internally deployed models.
Benchmarks & evals
Internal Eval Development
Evals for your own tasks, with grading criteria built with your team.
Model Upgrade Evaluation
Re-run the same tests on new releases and switch on evidence.
Evaluation as a Service
I run the evals for a defined workload and report, as an independent party.
Automation & Workflows
Custom workflows and pipelines for the processes no off-the-shelf tool covers. Deterministic where it can be, with an AI step only where judgement is needed.
Automation overview →Advisory, Deployment & Support
Every engagement in any area can start with planning and continue with support after launch.
Consulting & Planning
Stack evaluation, architecture and model selection, feasibility and build-vs-buy.
Uplift
Adding AI to processes and workflows you already run.
Deployment Options
Cloud, self-hosted or on-premises, matched to your security posture.
Ongoing Maintenance
Monitoring, updates and iteration for systems in production.
For industry-specific examples, see applications.
Two Ways to Engage
Every engagement includes hands-on building. The question is whether you also need help figuring out what to build.
Consult + Build
I help you figure out the right approach, then I build it. Ideal when you know the problem but need guidance on the solution — architecture, tooling, feasibility — before committing to implementation.
Consulting includes:
- Stack evaluation — audit your current tools and infrastructure for AI-readiness and improvement opportunities
- Stack component research — structured comparison reports when you need something new (models, platforms, digital wallets, vector stores, etc.)
- Architecture design & model selection
- Workflow audits & feasibility assessments
- Security posture & deployment planning
- Cost optimization & build-vs-buy analysis
- Custom model planning — training data, post-training methodology, fine-tuning providers & deployment
- Dataset blend & training recipe review
- Eval & benchmark development for model selection and upgrades
Then I build it:
- Agents, RAG, MCP integrations & private MCP servers
- Fine-tuned ASR, LLM and classifier models
- Workflows, data pipelines, dashboards & notifications
- Deployment, infrastructure & ongoing support
Just Build
You already know what you need — you have specs, architecture decisions, and a clear brief. I skip the discovery phase and go straight to implementation.
You provide:
- Defined requirements or spec
- Architecture / tooling decisions already made
- Clear scope and acceptance criteria
I deliver:
- Agents, RAG, MCP integrations & private MCP servers
- Fine-tuned ASR, LLM and classifier models
- Workflows, data pipelines, dashboards & notifications
- Deployment, infrastructure & ongoing support
Ready to get started?
Tell me about your project and I'll assess how I can help.