Generative AI development and DX consulting
Engineers who know the algorithms, building generative AI systems alongside you
Our people have supported AI and algorithm development with deep learning, mostly for manufacturers.
Solid engineering and honest project management, applied to a wide range of problems
Experience
Engineers who have been building AI systems since the early days of the field, and the quality of work that comes with it.
Value for money
An AI project turns on the people running it more than on the company name. Our costs sit at around a third of what AI system development usually runs to.
Maintenance
Selling is not where we stop. Talk to us about running it and improving it.
What we take on
01Generative AI and LLM development
Applying multimodal AI across text and images to your own systems — building the data, training the model, and developing the system around it.
- Document classification and management systems built on APIs
- LLM systems running standalone or on-premise, with a model we develop for you
02Generative AI and VLM development
Multimodal techniques across image and text, reading image and video more capably and more flexibly.
- Hazard-prediction systems working from image and video
- Image recognition systems that need no training data
03DX consulting
Consulting on renewing or building AI systems, from people who know the industry.
- Analysing why a visual inspection AI is not accurate enough (it does not have to be ours)
- Choosing AI software, and negotiating with the vendor
How a project runs
Free consultation
30 min – 1 hourWe hear out the inspection as it stands.
Talk to us.Pre-validation and quotation
1 – 2 weeksA short demonstration,
a proposal and an estimate.PoC
1 – 3 monthsValidating with AI whether the system holds together as a concept.
Requirements
up to 1 monthSettling what the software has to do in your environment.
Development and operation
several monthsBuilding the system to those requirements.
Case studies
Case 1: a consumer LLM system that holds a conversation like a person
We built the dataset, fine-tuned and reinforcement-trained an LLM, and developed a conversational model of their own
Point: two months from start to launch, supporting a few hundred thousand users
Case 2: helping a major Japanese manufacturer enter visual inspection AI
Matching them with the right partners, and preparing their material
Point: supporting a move out of hardware into AI, and well received for it
Case 3: rescuing an AI system already in place
The visual inspection AI they had adopted was poorly maintained; we supported it through to operation
Point: keeping what they had already paid for, and moving to operation at minimum cost
Case 4: help choosing an AI vendor
Standing on the buyer’s side as the AI specialist, so the negotiation is between equals
Point: a shorter project schedule and less system risk