Sathwik Tejaswi Madhusudhan

Apriel models

The Apriel-1.5-15B-Thinker paper (October 2025) lists Madhusudhan as technical co-lead for mid-training and post-training. The model starts from Pixtral-12B and applies a progressive three-stage training method.⁵³

Madhusudhan is the first named author of the December 2025 post introducing Apriel-1.6-15B-Thinker, which builds on Apriel-1.5 with a focus on text and vision reasoning and lower reasoning-token use. Madhusudhan also appears among the authors of the April 2026 Super Apriel paper.⁴⁶

The Apriel-H1 technical report (November 2025), on hybrid models that replace some attention layers of a 15B reasoning transformer with Mamba blocks through distillation, lists Madhusudhan as a technical co-lead under leads and management.⁷⁸

Madhusudhan's GitHub profile gives the role of machine learning engineer at ServiceNow. That profile is linked from the Hugging Face account that Hugging Face lists as an author of the Apriel-1.5-15B-Thinker paper and as the first author of the Apriel-1.6-15B-Thinker release post.¹²⁹⁴

Work and contributions

• Apriel-1.6-15B-Thinker Open weights

  First named author of the model's release post and listed under its leads and management.⁴

• Apriel model family Open weights

  Technical co-lead for mid-training and post-training on Apriel-1.5-15B-Thinker, as listed in the paper's contributions section.⁵³

Timeline

1. Oct 2025Served as technical co-lead for mid-training and post-training on Apriel-1.5-15B-Thinker.⁵³
2. Nov 2025Listed as a technical co-lead of the Apriel-H1 technical report.⁷⁸
3. Dec 9, 2025Published the Apriel-1.6-15B-Thinker release post as first named author.⁴
4. Apr 2026Co-authored the Super Apriel paper.⁶

Other documented roles

Roles the sources document that are not confirmed as current.

• Technical co-lead for mid-training and post-training on Apriel-1.5-15B-Thinker; listed under leads and management for Apriel-1.6-15B-Thinker (2025), ServiceNow³⁴

Where they work

• ServiceNow 6 people in the wiki

Related people

• Shruthan RadhakrishnaAlso at ServiceNow
• Torsten ScholakAlso at ServiceNow
• Raymond LiAlso at ServiceNow
• Oleksiy OstapenkoAlso at ServiceNow
• Paul SmithAlso at ServiceNow
• Victor SanhLanguage models · Multimodal models
• Sasha ShengLanguage models · Multimodal models
• Junyang LinLanguage models · Multimodal models
• Vaishaal ShankarMultimodal models · Language models
• Ilya SutskeverLanguage models · Multimodal models