Applied AI Engineer — Virtual Try-On (Contract, Part time, Remote)
Celier | Remote (India) | Contract
ABOUT
Celier builds an AI-powered virtual try-on mirror for physical clothing stores in India. A customer photographs themselves on an in-store touchscreen, selects a garment from the store's catalog, and sees an AI-generated image of themselves wearing it in 8–10 seconds.
THE ROLE
You will own the try-on pipeline end to end: person image + garment image → realistic composite, reliably, across Indian ethnic wear and Western clothing.
RESPONSIBILITIES
• Build and own the try-on inference service with a stable API
• Mask construction: human parsing, pose estimation, garment-agnostic region, face and hand exclusion
• Garment pre-processing at upload: background and mannequin removal, normalisation, caching
• Category-specific handling for sarees, lehengas, kurtas, sherwanis and Western garments
• A guardrail layer that automatically rejects broken outputs before display
• An evaluation harness scoring output quality across body types and garment types
• Latency and cost-per-generation optimisation
REQUIRED
• Production experience with diffusion-based image generation (not demos or notebooks)
• Inpainting and mask construction
• Segmentation and pose estimation: SAM, rembg, SCHP, OpenPose or DensePose
• Strong Python and PyTorch
• Ability to measure and systematically improve output quality
PREFERRED
• Prior virtual try-on work (IDM-VTON, CatVTON, Leffa, OOTDiffusion or similar)
• LoRA fine-tuning
• ComfyUI
• Familiarity with Indian garment types
NOT A FIT
LLM, RAG, chatbot or agent-focused profiles. This is an image generation role.
ENGAGEMENT
Part-time contract, remote, roughly 2–3 days per week. Begins with a paid, scoped 3–4 week engagement delivering a working pipeline plus a benchmark report. Ongoing contract thereafter. All code and configuration in a Celier-owned repository, running on Celier infrastructure.
TO APPLY
Send a link to relevant work — a rep
📌 Generative AI Engineer (India)
🏢 Celier
📍 India