Pavankumar Reddy Muddireddy

Timeline

1. 2017

   A zero-shot domain classifier

   In 2017, he co-authors the Interspeech paper "Zero-Shot Learning Across Heterogeneous Overlapping Domains" with Anjishnu Kumar, Markus Dreyer and Björn Hoffmeister. The text classification approach predicts which natural language understanding domain can handle a given utterance, including domains that did not exist at training time.

2. 2018

   Entity typing thesis at Illinois

   He completes an M.S. in electrical and computer engineering at the University of Illinois at Urbana-Champaign in 2018, advised by Dan Roth. His thesis, "Fine-grained entity typing system - design and analysis", introduces a modular architecture for fine-grained typing of entities.

3. 2021

   Updating a neural semantic parser

   In 2021, he co-authors "Overcoming Conflicting Data when Updating a Neural Semantic Parser" with David Gaddy, Alex Kouzemtchenko, Prateek Kolhar and Rushin Shah for the Workshop on Natural Language Processing for Conversational AI. The best method closes 86% of the accuracy gap between a naive data-mixing baseline and an oracle upper bound.

4. October 2024

   Co-authoring Pixtral 12B

   In October 2024, he is among the authors of Pixtral 12B, a 12-billion-parameter multimodal language model trained to understand both natural images and documents. The model surpasses a number of larger models on multimodal benchmarks.

5. June 2025

   Mistral's first reasoning model

   In June 2025, he is among the authors of Magistral, Mistral's first reasoning model. Magistral Medium is trained for reasoning on top of Mistral Medium 3 with reinforcement learning alone, and the open-source Magistral Small is released under Apache 2.0.

6. July 2025

   Co-authoring the Voxtral models

   In July 2025, he is among the authors of Voxtral, which presents Voxtral Mini and Voxtral Small, two multimodal audio chat models trained to comprehend both spoken audio and text documents while preserving strong text capabilities.

7. January 2026–March 2026

   Ministral 3 and Voxtral TTS

   In January 2026, he is among the authors of Ministral 3, a family of parameter-efficient dense language models in 3B, 8B and 14B sizes. In February 2026, he is among the authors of Voxtral Realtime, a natively streaming speech recognition model with sub-second latency. In March 2026, he is among the authors of Voxtral TTS, a multilingual text-to-speech model that generates natural speech from as little as 3 seconds of reference audio.

8. As of October 2026

   Audio research lead at Mistral

   He is a research scientist at Mistral AI and leads its audio research. In March 2026, as the Voxtral lead, he talks with the Latent Space podcast about the Voxtral TTS launch, and in September 2026, he joins Tim Scarfe on Machine Learning Street Talk for a technical tour of Voxtral.