Zahra Seyd

Product manager and computer vision engineer, owning AI and tech products end-to-end. Co-founder of correl8.ai, product observability platform for AI agents. Previously AI PM at Deepomatic (acquired by IQGeo) and CTO at AI.Nature (Inria).

Explore my work, side projects, and the parts of me that don't fit on a resume.

Hey, I'm Zahra, also known as Zar. I was born in Iran 35.8355° N, 50.9990° E and I'm — years old.

For the past 6 years I've been deep in the Paris startup ecosystem, building AI products alongside some of the best teams in Europe, first as an engineer, then as a product manager, and now as a founder. I started working on Correl8 because I got obsessed with that one question: what happens when AI agents become the new layer between users and products, and how that changes everything, not just the tech stack, but the UX, the analytics, the way teams make decisions.

Co-Founder & CEO Jul 2025 – Present

Correl8 AI · Paris

We're building the product analytics platform for teams shipping agentic products. We designed an MCP integration so agents report observations directly without exporting private conversations, and built a pipeline that clusters those observations into prioritized product, service, and agent issues with evidence and recommended actions. Piloted across 4+ use cases in consumer products.

AI Product Manager May 2022 – Dec 2024

Deepomatic · acquired by IQGeo · Paris

PM'd Deepomatic Studio end-to-end: a platform covering annotation, model training, deployment, evaluation, and continuous improvement. Also built the user-facing mobile app with my team, bringing computer vision models directly into field workflows.

CPTO Dec 2020 – Apr 2022

AI.Nature · Inria · Paris

Led engineering and product for a cross-functional team of engineers and biologists, taking computer vision models from research to a deployed product for biological applications.

Computer Vision Engineer Jan 2020 – Nov 2020

Vicomtech · San Sebastián

R&D on unsupervised and semi-supervised deep learning for noise reduction in medical imaging. Applied to fetoscopic laser photocoagulation surgery: denoised video frames and auto-detected vessels to assist surgeons in real time.

Rugswise ↗

Problem:

Iranian handmade rugs are one of the most intricate crafts in the world, and where a rug comes from determines most of its value. But the knowledge to identify provenance, weaving techniques, knot density, regional patterns, has always lived in the heads of a few experts, scattered and inaccessible. As an Iranian, I grew up around rugs and wanted to preserve both the craft and the knowledge around it.

Solution:

A provenance identification tool backed by a dataset I assembled of rugs from regions across Iran, annotated with weaving characteristics, material composition, and regional markers. The model takes an image and structured inputs and predicts origin, helping buyers, collectors, and inheritors understand what they actually have.

Joinifyouwant

Problem:

Always out of sync with friends for activities. We'd want to play badminton and need four people, but pinging everyone last-minute was painful enough that we'd just end up playing alone. And at some point we realised we didn't have time to see all our friends separately anymore.

Solution:

An app where anyone in your circle can post a last-minute activity and leave it open for others to join. You see what's happening, you show up if you want. It solved the team-forming problem and turned activities into a way to actually see friends, without scheduling a separate dinner two weeks out.

Mountains I've hiked or climbed.

2025 Tour du Mont Blanc 170 km

2025 Piton des Neiges 3,070 m

2026 Mont de Grange 2,432 m

2026 Le Linleu 1,711 m

2027 Damavand 5,610 m · coming