About FormaNova

The Vision

Imagine describing a piece of jewelry to an AI, or showing it an inspiring image, and receiving back a complete, production-ready 3D CAD model with individually editable components. Then imagine that same AI placing that design on a virtual model's hand, generating a photorealistic product image, and pushing it directly into a Shopify store. Finally, imagine using that same CAD file to order manufacturing — but only when demand exists. No inventory, no risk.

This isn't science fiction. This is FormaNova.

We're building the end-to-end automation layer for jewelry design, production, and commerce.

Our mission is simple: remove the gatekeeping that has locked jewelry design behind expertise, capital, and inventory constraints. We want to enable everyone to be a jewelry designer, power on-demand manufacturing at scale, and usher in an era of personalization the jewelry industry has never seen.

The Problem

Jewelry is one of the highest-margin industries in the world (50–300% retail margins). But it's also one of the most gatekept.

Designing requires CAD expertise combined with deep knowledge of jewelry construction — skills that take years to develop. Manufacturing requires capital and inventory risk. Customization is slow and expensive; the design phase involves lengthy feedback cycles and render iterations. A handful of luxury brands control the entire narrative because they can afford the design, photoshoot, and manufacturing infrastructure.

Meanwhile, generative AI has transformed every creative field. Except jewelry.

Why? Because jewelry isn't just hard to design; it's hard to represent computationally. Take a ring for example. It isn't a single, solid object — it's an assembly: band, prongs, bridge, settings, stones. Each component needs to be separately editable. A diamond's light refraction needs to be photorealistic. A design needs to look identical to the input, without hallucination.

The existing AI tools — standard diffusion models, off-the-shelf 3D generators, generalist photography engines — all fail at jewelry because they treat it like everything else. We saw this gap and decided to go deep.

Our Approach

Jewelry-first specialization. We are not building a generalist AI tool that happens to work on jewelry. We are building for jewelry, with jewelry-specific research, training data, and model architectures.

Discrete mesh generation, not blobby geometry. When most AI systems generate 3D objects, they create a single fused mesh — like a melted candle. Jewelry designers hate this because you can't edit individual components. FormaNova generates jewelry as a kit of parts: each prong, each stone, each segment of the band exists as a separate, transformable object. Users get full object-level control in the CAD editor before export.

Domain expertise over generic models. Our team brings together applied AI researchers, physics graduates who understand light transport, and jewelry industry specialists who know what actually works in manufacturing. We've spent over a year curating datasets, fine-tuning models, and building architectural innovations that generalist models will never prioritize.

Photography fidelity at scale. Beyond CAD generation, FormaNova also solves the immediate photography bottleneck: taking jewelry images as input and generating photorealistic product shots on virtual models. FormaNova's QA evaluation of 316 actual user generations found 82.9% required no regeneration on first generation — a best-in-class one-shot accuracy rate, keeping users out of generation hell.

Research Foundation

Our approach is grounded in rigorous research. FormaNova has published 76 open-source, gold-standard datasets on Hugging Face (raresense), spanning virtual try-on, inpainting, upscaling, and jewelry-specific training data. These datasets are publicly accessible (gated) and represent thousands of hours of human curation and annotation.

This research library serves two purposes:

  1. It enables our own model training and architectural innovations
  2. It provides research artifacts that other teams in the community can build on

The Founders

Sophia Pervez and Hassan Baig, co-founders of FormaNova

Sophia Pervez, Co-founder & Chief Revenue Officer
Sophia brings institutional rigor from IBM, combined with battle-hardened startup experience and deep connections in the NY jewelry and fashion executive community. She's a mentor at the Women's Jewelry Association and leads all go-to-market strategy, partnerships, and revenue. Her mission: to revolutionize jewelry design and commerce.

Hassan Baig, Co-founder & Head of Product & Technology
Hassan is a Duke graduate and serial founder who has built consumer products at significant scale (40M+ users at peak). He's self-taught in engineering, moving from non-coder to full-stack architect. When image generation emerged, he immediately recognized its potential to transform visual commerce. At FormaNova, he's the technical architect behind core innovations — from discrete mesh generation to proprietary model fine-tuning. His North Star is ensuring FormaNova's technology translates into real customer value and market traction.

The Future of Jewelry

FormaNova: a fully automated pipeline from imagination to product to customer, constrained only by creativity, not expertise. We believe this represents a fundamental shift in how jewelry gets designed, produced, and sold. Not a feature. A category.

FormaNova is built by Rare Sense, an applied AI research lab based in New York City.