Sophia Yang

Sophia Yang is an AI developer relations leader and educator who joined Fireworks AI to lead developer relations, announcing the move on September 15, 2026. Previously Mistral’s head of developer relations, she helped build its DevRel team, demonstrated practical model applications, and coauthored Mixtral of Experts. She publishes under the handle sophiamyang.[3][4][1][6][2]

In her Fireworks announcement, Yang described her goal as “owning your own intelligence.” For her, that means going beyond running open models to making them specialized through a developer’s data, evaluations, and feedback loop. She joined an existing developer relations team to pursue that mission.[3]ⁿ¹

In her August 21, 2026 career-transition post, Yang said she was between jobs and had committed to a new role. Her embedded Mistral farewell described leaving as an especially difficult career decision—harder than her earlier move to France without speaking French. She looked back on building Mistral’s developer relations team, crediting her teammates for their shared work and her founders, managers, and advisors for their support.[4]ⁿ²

Before that transition, Yang was one of the 26 coauthors of Mixtral of Experts, submitted on January 8, 2024, alongside first author Albert Q. Jiang and other Mistral collaborators. The paper describes sparse mixture-of-experts routing: each layer contains eight feedforward experts, and a router selects two for each token and combines their outputs. The authors report that each token has access to 47 billion parameters while using 13 billion active parameters during inference. Both the base and instruction-following models were released under Apache 2.0. The collective authorship credit does not specify Yang’s individual technical responsibilities.[6]

At Latent Space LIVE at NeurIPS 2024, Yang made model capabilities concrete through interactive demonstrations. In the presentation published December 23, she asked Mistral’s Le Chat to read a receipt, isolate coffee and sausage charges, and add an 18 percent tip. She also demonstrated a canvas using PyScript to execute Python in the browser and showed generated Tetris code that users could ask the interface to explain or modify. These were practical demonstrations rather than general accuracy or reliability measurements. Latent Space identified her as Mistral’s head of developer relations and also noted her presentation at the AI Engineer World’s Fair Open Models track.[1]

Yang’s teaching work predates those AI demonstrations. On October 17, 2022, Anaconda introduced her as its resident senior data scientist and instructor of Introduction to Data Visualization with Python. Her course taught learners how to approach and understand data and derive insights, providing an earlier example of her work explaining technical tools through instruction.[7]

Her public LangChain tutorial repository also offers overview and question-answering notebooks, plus a Panel application notebook. Its accompanying material addresses chatting with long PDF documents and multiple approaches to question answering, giving developers concrete educational artifacts to explore.[5]