Theodore
Helped ship Neuono, the AI-led fashion label that launched at London Tech Week 2025, in a one-month sprint from idea to app.
- Role
- AI Project Development Intern
- Employment type
- Internship
- Year
- 2025
- Team
- One of 8 software interns, reporting to the co-founder
- Company
- THDR Group (Theodore / Neuono)
What would you like designed?
Theodore is a Sydney tailor for men and women, part of THDR Group, which pairs generative AI with made-to-measure craft, so nothing is made until someone needs it. Neuono, its AI fashion app, puts that tailoring in a phone: live on iOS and Android, it turns a few words and a selfie into an outfit Theodore can make.
- Made to fit
- Personalised
- Tailored, not mass‑manufactured
Delivered& how it works
A selfie and a prompt
Upload a selfie, describe the outfit: “red dress, beach wedding”.
Measured on the phone
Pocket Tailor measures the body with the phone’s own AR sensors, so every outfit is made to fit.
Built by THDR before Neuono: 97.5% accurate across 20,000+ scans, by THDR’s figures.
Reading the request
Through the OpenAI API, the prompt is parsed into structured fields and expanded into a generation prompt.
I was part of the group that brainstormed the whole system architecture.
My part
200+ LoRAs, one for each Theodore design
<2 min for the whole workflow
SenseThread is Neuono’s generation pipeline, built in Retool on GCP with several model providers. My part was the image generation itself: Stable Diffusion in ComfyUI, backed by AlloyDB. I gathered training data for each of Theodore’s 200+ designs, across styles, lengths, colours, lapels, collars and cuts for menswear and womenswear, then trained and expanded a dedicated LoRA for every design, so each result is a garment Theodore can actually tailor. I benchmarked model providers and worked with the team to bring the full workflow under two minutes.
My part
Three views
One outfit, seen from the front and the back, on a model who resembles the user. Diffusion models generate every image from scratch, so fabric shifts between views and details vanish or appear. I built a multi-view ComfyUI workflow that keeps the garment consistent across angles, and integrated fal.ai’s image-generation API.
In the press
10,000+ Neuono downloads in its first three months
700+ trillion possible garment combinations, by THDR’s count
Throughout the internship, they demonstrated an exceptional blend of technical expertise, creativity, and dedication. … Their ability to integrate complex technical solutions with creative design insights was particularly impressive and greatly enhanced the overall success of our project.
Stack: Python, JavaScript, SQL, OpenAI API, Stable Diffusion, ComfyUI, LoRA fine-tuning, fal.ai, Computer vision, RAG, Model evaluation, Retool, FlutterFlow, PostgreSQL / AlloyDB, Firebase, Google Cloud