Md Ashikur Rahman

• Leading early-stage R&D for CogniX, a proprietary multimodal image-generation and editing initiative for product visualization workflows, aimed at reducing reliance on third-party model providers.
• Exploring multi-reference fusion and prompt-guided image editing for garment replacement, scene composition, product visualization, and style transfer using LoRA-based adaptation, quantized fine-tuning experiments, PyTorch/Diffusers, and ComfyUI workflows.
• Investigating dataset design for variable-length multi-reference inputs, semantic instructions, attribute-rich prompts, and preservation constraints.

• Conceived and co-led a unified cross-brand virtual try-on platform across 170+ brands.
• Designed the user workflow where shoppers can browse products from hundreds of brands, virtually try them on, and purchase from one place.
• Architected Dockerized FastAPI/Nginx backend with SQL Server, Google Cloud Storage, Redis, recommendation workflows, and 2D try-on pipelines.

• Leading AI catalog-audit and image-QA platform for a major US retail client, automating image QA, catalog validation, metadata checks, and content-health monitoring across DAM/CMS workflows.
• Built image-audit workflows to detect inappropriate content, sheer-garment visibility, offensive tattoos, vulgar copy, crop inconsistencies, retouching defects, face-crop issues, and duplicate or near-duplicate images.
• Developed data-audit capabilities to validate product copy, color and style attribution, product-type metadata, image origination tags, and DAM/CMS integration requirements.
• Designed backend audit checks for inventory gaps, broken image and swatch links, alt-text quality, and pages requiring updates or removal.

• Co-designed architecture and production APIs for prompt-based image and video generation with multi-model selection, reference-image conditioning, background processing, and Llama-based prompt classification, intent routing, and model-selection workflows; onboarded 200+ brands for creative and product-image workflows.
• Built production APIs and backend workflows for generation requests, background processing, and delivery to downstream systems.
• Enabled users to select from multiple state-of-the-art generation models for image and video creative workflows.

• Designed a multi-view 3D reconstruction pipeline for generating textured GLB/OBJ assets from 360° video and image sequences, combining COLMAP-based camera pose estimation and structure-from-motion, multi-view stereo reconstruction, Open3D mesh processing, and experimental NeRF/3D Gaussian Splatting refinement.
• Optimized reconstruction quality and processing efficiency to reduce manual 3D cleanup and editing effort.

• Developed production salient-object segmentation for background removal; improved segmentation quality by 17% on internal benchmarks, reduced processing time by 30%, supported uploads up to 257 MB, and achieved 2.27s average processing time across standard production workloads.
• Scaled Retouched.ai to 4.5M+ images processed globally across hundreds of customers using PyTorch, U²-Net-inspired salient-object segmentation architecture, FastAPI, and GCP.

Production client: leading European luxury resale marketplace

• Optimized .NET/SQL Server workflow for a leading European luxury resale marketplace; integrated Omnimage.ai and Retouched.ai across FTP ingestion, QC validation, and automated delivery; ~17K images/day, ~65-minute average batch turnaround.

• Built customer-service QA workflow with a SeamlessM4T-based speech transcription/translation pipeline and Llama 2-based contextual key-phrase extraction; automated transcription, validation, and review pipelines via FastAPI, reducing manual audio review effort.