01
AI product direction
Feasibility, model selection, system boundaries, risk thinking and a practical plan for where AI genuinely belongs in the product.
Applied AI strategy, prototyping and systems
Useful AI for products, workflows and private business systems.
This offer is for businesses that want practical AI work rather than a slide deck. It covers AI features inside apps, internal copilots and knowledge tools, local and private model setups, and the architecture needed to make those systems dependable.
What this work looks like
AI inside real products, internal tools and local-first knowledge systems.
Model choice, tool design, retrieval, prompt strategy, evaluation and the UX needed to make AI feel useful instead of chaotic.
MCP servers
RAG workflows
Local LLMs
App integrations
Tooling familiarity
Frontier APIs and local model stacks.
OpenAI, Claude, Gemini, xAI, Ollama, LM Studio and open-source tooling.
Already in use
TimerStack and PoeticAI both ship AI-backed product features.
Studio stance
Private where needed. Product-led where it matters.
AI product strategy
Model selection
MCP workflows
RAG systems
Private deployments
App integrations
AI product strategy
Model selection
MCP workflows
RAG systems
Private deployments
App integrations
Services
Four ways businesses usually need AI help.
The value is usually not "an AI app" in the abstract. It is a better internal workflow, a smarter product feature, a private knowledge system or a clearer technical direction before money gets wasted.
02
AI inside apps
Content generation, assistant flows, prompt and tool design, structured outputs and product UI that keeps the AI useful to the end user.
03
Private knowledge systems
RAG and document workflows for teams that need their own material organised, searchable and usable without shipping everything to a public SaaS layer.
04
Local and networked AI setups
Ollama, LM Studio and local model serving for businesses that care about privacy, latency, offline access or keeping experimentation inside their own environment.
Relevant experience
Broad AI familiarity is only useful if it turns into practical judgement.
The point of this offering is not allegiance to one provider. It is enough hands-on experience across frontier models, local tooling and product integration to recommend what is actually fit for the business.
Early OpenAI API work
Worked with the OpenAI API during its beta period and provided early feedback while the platform was still taking shape.
Provider and tooling breadth
Actively reviewed OpenAI, Claude, xAI, Gemini, Hugging Face model options and local stacks such as Ollama, LM Studio and related tooling.
Agent and integration systems
Built MCP infrastructure, including a server for serving TimerStack data on the web, and worked with skills and workflow-driven tool chains.
Private retrieval and document systems
Created a local RAG database over thousands of files for sensitive personal research, and used tools such as NotebookLM where the job called for them.
Local deployment experience
Set up machines to serve locally hosted LLMs across a network for cases where private infrastructure and direct control mattered.
AI shipped in real products
TimerStack and PoeticAI both use AI to generate useful content inside the product rather than treating AI as decorative marketing.
Engagements
How businesses can buy this work.
AI is easiest to sell when the engagement shape is obvious. These are the three cleanest ways to package it without drifting into vague consultancy.
Advisory and feasibility
Short engagements for leadership teams that need help deciding what to build, which models to trust, where the risks are and what the first useful version should be.
Prototype and pilot
Focused build work to prove an assistant, content generator, knowledge system or internal workflow before a larger budget gets committed.
Implementation and retained support
Ongoing technical involvement for teams that need integration, iteration, evaluation, operational hardening and a responsible owner for the AI layer.
Good fit
Businesses with proprietary documents, repeatable internal workflows, product ideas that genuinely benefit from AI, or teams that need a grounded second opinion before committing to a vendor stack.
Not the offer
Generic prompt workshops, one-click automation theatre or AI positioning work with no operational use behind it. The work needs to end in a real system, a sharper product or a clearer decision.
Contact
If you have an AI idea, internal workflow or private knowledge problem worth shaping properly, let's talk.
The best starting point is usually a conversation about the business problem itself, the data involved and whether the answer is product integration, local tooling, retrieval, agent workflows or a simpler non-AI fix.