Cyrine Chaabani

AI Research & Engineering

Cyrine Chaabani

I am a Co-founder & CTO at Wattzo, building AI-powered energy optimization platforms. Prior to this, I worked as an Applied Scientist at Microsoft, focusing on speech and sign language detection, and as a Founding Lead AI/ML Engineer at Xaver. I hold a Master's degree in Robotics, Cognition, and Intelligence from the Technical University of Munich and an Honors Degree in Technology Management from CDTM.

My research and work leverage foundation models for solving complex tasks through specialization and reinforcement learning. Beyond this, I am interested in knowledge-representation, online learning and graph neural networks.

Always feel free to reach out with things you find exciting.

Publications

1. Explainable Energy Communities with Generative AI
IEEE ISGT, 2024
Introduces a generative-AI interface for Renewable Energy Communities that enables natural-language queries over optimization and forecasting workflows. Demonstrates improved transparency and decision support in case studies of community energy management.
2. Signer Overlap in Sign Language Detection
Preprint, 2023
Shows that training/test signer overlap inflates performance in sign-language detection. Proposes non-overlapping splits and reports a measurable drop when overlap is removed, arguing for more realistic evaluation protocols.
3. Clean Text & Full-Body Transformer for Sign Language Translation
WMT22, 2022
Describes a system that couples full-body visual features with careful target-text cleaning to address large vocabulary and data scarcity in SLT. Achieves state-of-the-art results in the WMT'22 task with strong human evaluation.

Projects

Blog Summarizer MIT Sundai Club
04/25
Built Blog to Talking Head Video Generator. This repo transforms blog post content into engaging short videos featuring an AI talking head.
GenAI UX Best Practices BMW
10/24 - 02/25
Researched and defined UX/UI best practices for GenAI products and tools to boost their integration and ease-of-use for all categories of employees and for all use-cases. The efficacity of the then defined approach was proven through user-experiments and feedback.
Unbiased Cathode MIT Energy and Climate Hackathon
11/24
LLM-based co-pilot for battery material researchers that helps them quickly understand the supply-chain constraints (E, S and G) of the materials they are using in their experiments.
Konvolut CDTM
10/22 - 02/23
Built an ESG reporting co-pilot and early supply chain risk detector for SMEs. It relies on sentiment analysis from specialized data sources to detect any warning signs from known suppliers, as well as suggest potential alternatives. Additionally, it generates compliance reports seamlessly.

Education

Honors Degree, Technology Management, Center for Digital Technology and Management (CDTM)
2022 - 2024
Focus: product management, rapid prototyping, leadership
M.Sc. Robotics, Cognition, Intelligence, Technical University of Munich
2019 - 2023
Focus: deep learning, NLP, generative AI, graph learning, robotic motion planning
B.Sc. Electrical Engineering & Information Technology, Technical University of Munich
2016 - 2019
Focus: signal processing, control systems, embedded systems, computationl Intelligence

Volunteering

Co-founder and Impact lead of the Munich Hub, building AI projects for the public good
Member of the leadership team of the Munich Hub and co-founder of the Climate Café
As member of the leadership team, co-organized and led the weekly sessions of MUNTUM and co-organized and participated in local MUN conferences.

Awards

1st Prize Winner @ MIT Energy & Climate Hackathon
11/23
Best Model for Sign Language Translation @ WMT22
06/22
Academic Excellence Scholarship
09/15 - 09/22 (0.04% acceptance rate)

Languages

Arabic C2
French C2
English C1
German C1