Laurent Mazare

Paris, Ile-de-France

Join Prog.AI to see contacts

Summary

🤩

Rockstar

Laurent Mazare is a pragmatic CTO based in Paris with 13 years of hands-on experience building backend systems and production-ready ML infrastructure. He leads Kyutai’s technical strategy while contributing deeply to open-source projects, notably developing core features for Moshi, a streaming speech-text foundation model and full‑duplex dialogue framework. Comfortable in Rust and ML engineering, Laurent has improved compiler lints in rust-clippy and added KV cache support to the minimalist Candle ML framework, reflecting a focus on performance, correctness, and efficient model decoding. He combines systems-level rigor with research-adjacent work on quantization, device orchestration, and streaming audio codecs, making him as fluent in low-level optimization as in shaping product-grade AI platforms.

14 years of coding experience

Github Skills (25)

pytorch10

system-configuration10

python10

boolean-logic10

machine-learning10

inference10

custom-configuration10

trainings10

lint10

model-optimization10

linter10

configurations10

rust10

yml-configuration10

modeling10

Programming languages (11)

C++ShellRustCOCamlSwiftHTMLMetal

Github contributions (5)

huggingface/candle

Jun 2023 - Jun 2025

Minimalist ML framework for Rust

Role in this project:

Back-end Developer & ML Engineer

Contributions:631 reviews, 2789 PRs, 3440 pushes in 1 year 11 months

Contributions summary:Laurent's commits focus on the implementation of a KV cache to the T5 model within the "candle" repository. Their work involved adding and integrating a KV cache for decoding, suggesting the use of release mode, and adding comments. The primary focus of their work is on the minimal ML framework for Rust, which indicates this user is likely involved in developing the framework's core backend and is involved with machine learning model optimization.

machine-learningrust

kyutai-labs/moshi

Sep 2024 - Jan 2026

Moshi is a speech-text foundation model and full-duplex spoken dialogue framework. It uses Mimi, a state-of-the-art streaming neural audio codec.

Role in this project:

Back-end Developer

Contributions:3 releases, 47 reviews, 168 PRs in 1 year 3 months

Contributions summary:Laurent primarily contributed to the core functionalities of a speech-text foundation model and spoken dialogue framework, moshi. The commits demonstrate the user's involvement in configuring and instantiating language models (LMs), integrating them with Encodec models. They also focused on adjusting quantization configurations and implementing proper device handling and various testing and generation workflows demonstrating expertise in machine learning, model training and the supporting infrastructure.

audiocodecfull-duplexmoshispeech

Find and Hire Top DevelopersWe’ve analyzed the programming source code of over 60 million software developers on GitHub and scored them by 50,000 skills. Sign-up on Prog,AI to search for software developers.

Request Free Trial