01 — What Myara does STATUS // IN DEVELOPMENT

An AI tool for pattern cutting, built around the fabric you actually cut.

Myara is being built to take the repetitive work out of the pattern room. First to go live is image to pattern: a vision pipeline reads the garment and an ontology of pattern-making knowledge drafts it, so every line follows the same rules a pattern cutter would. The result is a CAD-ready DXF that drops into the workflow you already use.

  1. 0

    Image to pattern

    A vision pipeline reads the garment from an image; the ontology drafts the pattern pieces, with notches and grain lines.

    First to go live
  2. A

    Automated grading

    A full size range from one base pattern, graded by rule instead of by hand.

    In development
  3. B

    Automated pattern adjustments

    Alterations to length, width, balance and fit carried consistently through every connected piece.

    In development
  4. C

    Tailoring / made-to-measure

    A pattern drafted to one person's body measurements, not only to a size chart.

    In development
  5. D

    Fabric-informed modification

    Ease and shaping adapted to a specific fabric and its weight, so the same garment cuts right in a light wool or a heavy coating.

    In development
Input
Garment image · base pattern · body measurements · fabric and weight
Knowledge
Vision pipeline + ontology of pattern-making rules
Output
CAD-ready DXF
Status
In development
02 — Why ORIGIN // LONDON 2021 — 2026

It started with a pile of scraps.

Myara began in a London student bedroom. Studying fashion design, our founder preferred draping to drafting: pinning fabric on a mannequin, cutting into it, learning its weight and stretch by hand. Every offcut felt too useful to throw away, and within months a corner of the room was scraps.

That was one student. Multiply it by every fashion school and an industry that turns collections around in days, and the waste becomes hard to look at.

The tools meant to help were slow to learn. 3D software like CLO took time away from the fabric itself. Fashion manufacturing has room to modernise, on one condition: new technology has to fit into the workflows pattern rooms already trust.

Footnote — tailoring runs in the family: a grandfather who was a tailor in Poland.

  1. 2021 — 22London. Draping over drafting; the scraps pile up.
  2. 2023Berlin placement. First experiments with generative AI.
  3. 2024BA (Hons) Fashion Design & Development, First Class.
  4. 2024 — 25A year of code, mathematics and web development.
  5. 2025 — 26MSc AI, UCL. Generative pattern geometry from images.
  6. NowMyara.
03 — Case study ACTIVE RESEARCH // NOT A PRODUCT FEATURE

Active research

Further down the line: image to pattern with a generative diffusion model.

Myara's first release reads garment images with a vision pipeline and drafts with the ontology. Further down the line, we are researching a generative approach: a diffusion model, trained on professionally drafted industrial patterns, that produces 2D pattern geometry directly from a garment image.

The hard part is validity: the model has to output pattern pieces that are geometrically well-formed. Across the research runs so far, valid output rose from 54% to 99.4%, and stayed at 96% on held-out classes of the same garment type. Generalising to other kinds of garment will need much more data.

This is ongoing research, not something Myara offers today. The numbers come from research runs and may change as the work continues.

Garment image diffusion model 2D pattern geometry

  • Valid output — first runs 54%
  • Valid output — latest runs 99.4%
  • Held-out classes, same garment type 96%

5-fold cross-validation · professionally drafted industrial data · MSc research, UCL

04 — Founder AUTHOR // MLK
Portrait of Marcel L. Kedziora
MLK // FOUNDERLONDON

Marcel L. Kedziora

AI engineer and researcher in geometry, supply chain and manufacturing for fashion. Trained first in pattern cutting, draping and garment construction, Marcel moved into AI to fix what he saw on the cutting table. He has led research for a London-based CAD start-up, consulted on an international R&D programme for AI in fashion, and founded Sustainable AI Solutions Ltd. He is an invited speaker on AI in fashion at London College of Fashion and the China Academy of Art.

Research
Research lead — London-based CAD start-up
Consulting
International R&D programme, AI in fashion
Talks
Invited speaker, AI in fashion — London College of Fashion · China Academy of Art
MSc
Artificial Intelligence — UCL
BA (Hons)
Fashion Design & Development, First Class — UAL: London College of Fashion
Company
Sustainable AI Solutions Ltd — Founder
Languages
English · French · Polish
Base
London, United Kingdom
05 — Writing & talks DOC 01 / 01

Invited talk — China Academy of Art

CAD AI in Fashion: Automating Pattern Cutting

“I came into AI from fashion design, initially bothered by the inefficiency of traditional iterative sampling. I found it wasteful and draining, and the waste I was creating for my studies was piling up in my London bedroom. We all know that is far worse at scale.”

— Marcel L. Kedziora
Read the full text

I am an AI engineer and researcher specialising in geometry, supply chain and innovative manufacturing methods in the fashion industry, building AI engines that have the potential to automate processes and reduce waste at scale.

I came into AI from fashion design, initially bothered by the inefficiency of traditional iterative sampling. I found it wasteful and draining, and the waste I was creating for my studies was piling up in my London bedroom. We all know that is far worse at scale.

Through my AI studies and professional work, I also realised that the fashion industry is slow to adopt new technologies, and I was part of that too. Learning 3D software like CLO took time I didn’t have, time I would rather have spent touching real fabric and understanding its weight and stretch, designing from the fabric rather than from the drawing. Many professionals share this experience, and in many places manual work remains cheaper than adopting new tools.

Comparing fashion with automotive, defence, architecture and other engineering disciplines showed me how much room there is to modernise fashion manufacturing. It also made it clear that new technologies must integrate seamlessly into established workflows rather than ask people to change everything. Some steps will take years to automate, but others can be automated now, and even a 10% reduction in lead time is massive at scale.

I am currently working on automating pattern cutting by combining a vision model pipeline with an ontology that encodes pattern-making knowledge. Encoding all of it takes time, so I am starting with specific garments, with the capacity to expand to others depending on the needs of factories and brands. Over time, these processes can reduce waste and cost and, in the longer term, make tailored garments accessible to everyone, as pattern alterations become natural at scale. During the presentation, I aim to show what I have been building so far, its accuracy, and how this technology can be integrated and developed further.

I believe agentic workflows are very promising for building these pipelines, in pattern cutting and in other parts of the fashion industry. What the industry needs most is openness about its pain points, so that researchers can help solve them and create growth through less waste worldwide.

I would like this presentation to inspire and open conversations with emerging and established professionals, manufacturers and brands, to demystify some of the marketing hype around AI, and to encourage everyone to experiment with agentic workflows and coding tools to speed up their work and reduce their own waste.

  • China Academy of ArtInvited speaker — CAD AI in Fashion: Automating Pattern Cutting
  • London College of FashionInvited speaker — AI in Fashion, November 2026