01
Foundation
Understand products
Create accurate, trustworthy product understanding from descriptions, images, attributes and other signals.
Boutique lab for retail
Understand products. Enrich experiences. Transform retail.
Tinkerkraft builds AI systems that turn complex product data and catalogues into connected knowledge—powering better decisions and more relevant customer experiences.
01 Perspective
A catalogue records what you sell. A knowledge system understands it.
Retail catalogues were built to store products, not to understand them. The information is fragmented, inconsistent and rarely captures all the meaning that customers and retail teams need.
We build the systems that turn this complexity into knowledge that is structured, connected and ready to use.
What we believe
Better AI starts with the right context.
Which attributes matter? When are two products meaningfully equivalent? What information can be trusted? These are domain questions—not just modelling problems.
02 Expertise
From product data to intelligent retail.
One connected journey—from making sense of products to creating experiences that understand what people are looking for.
02
Structure
Connect knowledge
Model the relationships between products, categories, catalogues and the concepts that give them meaning.
03
Experience
Enrich experiences
Apply deeper product understanding to improve how products are managed, discovered, compared and chosen.
In practice Problems we solve
Classification at scale
Placing varied products into complex, evolving taxonomies while handling uncertainty and exceptions.
Product matching
Recognising when incomplete and inconsistent listings refer to the same product across catalogues.
Catalogue quality
Measuring whether product information is complete, consistent, trustworthy and fit for purpose.
Retail semantics
Designing taxonomies, ontologies and knowledge structures that capture how products actually relate.
Claims and trust
Applying AI to nuanced product claims where language, evidence and context all matter.
Discovery and relevance
Connecting product meaning with customer intent to make finding and choosing feel more natural.
03 Approach
Built for retail’s messy reality
First principles.Useful systems.
We break challenging problems down to their fundamentals—what must be true, what evidence exists and how success should be measured—then build the simplest system that can work.
- 01
Start from first principles
Question assumptions, make the fundamentals explicit and frame the real problem before choosing an approach.
- 02
Model the knowledge
Make the domain, relationships, language and sources of truth explicit.
- 03
Experiment and evaluate
Define what good means, test honestly and understand the consequences of different errors.
- 04
Build for use
Develop systems that work with people, adapt to change and improve over time.
Principles How we think
Domain before models
We begin with the domain—its language, constraints and sources of truth. The technology follows.
Driven by data
We interrogate the data to uncover patterns, challenge assumptions and ground design decisions in evidence.
Outcomes over outputs
We measure success by decisions improved and value created—not features shipped or models deployed.
Value in every iteration
We work in small, testable increments—putting useful capability into practice early and learning what to build next.
Trust through evidence
We make uncertainty and failure modes visible, measure quality honestly and bring in expert review when judgement is required.
Technology without dogma
LLMs, ontologies, knowledge graphs and agentic AI are building blocks. We use what the problem demands.
04 Careers
Join Tinkerkraft
Help machines understand the world of products.
We’re building a small, thoughtful team for people who enjoy hard problems, intellectual honesty and the responsibility to turn promising ideas into useful systems.
Send us a short introduction and something you’ve built or solved.
Introduce yourselfWe’d like to meet
- 01Product-minded problem solvers
- 02Applied scientists
- 03AI developers
- 04Data & knowledge engineers
05 Contact
What could deeperproduct understandingunlock for your business?
Bring us a difficult catalogue, discovery or retail-knowledge problem.
ai@tinkerkraft.com