AC

Arnav Chavan

Senior Applied Scientist, Amazon Lab126 · Incoming Ph.D. Student, UC Berkeley

Research Interests

Neural Dynamical Systems

Treating computation as continuous flow — ODEs & SDEs as a design language for efficient, expressive models.

AI for Scientific Discovery

Turning those same dynamics outward — models that surface structure and accelerate discovery in the sciences.

Continual Learning

Learning as a trajectory with no endpoint — absorbing new tasks over time without overwriting the past.

Robotics on the Edge

Bringing real-time perception and control to robots under tight on-device compute, memory, and power budgets.

About

I'm a Senior Applied Scientist at Amazon Lab126, where I work on extreme hardware–software co-design — leading generative-AI capabilities on Amazon's custom silicon to bring on-device GenAI to products like Alexa+. I joined through the acquihire of Nyun AI, the startup I co-founded and led as CTO.

In Fall 2026 I'll begin a Ph.D. at UC Berkeley. I'm increasingly drawn to efficient architectures through the lens of Neural ODEs/SDEs, AI for scientific discovery, and continual learning — building on a track record in model compression, quantization, and efficient foundation models — with papers at CVPR, ICML, ICLR, IJCAI, ACL, and TMLR.

Earlier, I was a co-founding member of Transmute AI Research, and a researcher at Microsoft Research, Google DeepMind, and MBZUAI.

News

Selected Publications

* equal contribution. Full list on Google Scholar.

2026
S2D: Selective Spectral Decay for Quantization-Friendly Conditioning of Neural Activations
Arnav Chavan*, Nahush Lele*, Udbhav Bamba, Sankalp Dayal, Aditi Raghunathan, Deepak Gupta
CVPR 2026arXiv ↗
DOT-MoE: Differentiable Optimal Transport for MoEfication
Udbhav Bamba*, Arnav Chavan*, Aryamaan Thakur, Steve Teig, Deepak Gupta
ICML 2026arXiv ↗
Jacobian-guided Noise Injection for Quantization Robustness in Large Language Models
Deepanshu Pandey*, Nahush Lele*, Arnav Chavan*, Sankalp Dayal, Deepak Gupta
ICML 2026 Workshop
2025
Rethinking the Value of Training-Free Structured Pruning of LLMs
Arnav Chavan*, Nahush Lele*, Aryamaan Thakur, Deepak Gupta
One-for-All: Generalized LoRA for Parameter-Efficient Fine-tuning
Arnav Chavan*, Zhuang Liu, Deepak Gupta, Eric Xing, Zhiqiang Shen*
PreprintarXiv ↗
2024
Surgical Feature-Space Decomposition of LLMs: Why, When and How?
Arnav Chavan, Nahush Lele, Deepak Gupta
ACL 2024Paper ↗
Faster and Lighter LLMs: A Survey on Current Challenges and Way Forward
Arnav Chavan, Raghav Magazine, Shubham Kushwaha, Mérouane Debbah, Deepak Gupta
IJCAI 2024Paper ↗
Pushing the Limits of Gradient Descent for Efficient Learning on Large Images
Arnav Chavan*, Deepak K. Gupta*, Gowreesh Mago*, Dilip K. Prasad, Rajat Thomas
Rethinking Compression: Reduced Order Modelling of Latent Features in LLMs
Arnav Chavan, Nahush Lele, Deepak Gupta
Beyond Uniform Scaling: Exploring Depth Heterogeneity in Neural Architectures
Arnav Chavan*, Akash Guna R.T*, Deepak Gupta
2023 & earlier
A Comparative Study of Model Compression Techniques on Fairness in Language Models
Krithika Ramesh, Arnav Chavan, Shrey Pandit, Sunayana Sitaram
ACL 2023Paper ↗
Vision Transformer Slimming: Multi-Dimension Searching in Continuous Optimization Space
Arnav Chavan*, Zhiqiang Shen*, Zhuang Liu, Zechun Liu, Kwang-Ting Cheng, Eric Xing
CVPR 2022Paper ↗
Dynamic Kernel Selection for Improved Generalization and Memory Efficiency in Meta-learning
Arnav Chavan*, Rishabh Tiwari*, Udbhav Bamba, Deepak K. Gupta
CVPR 2022Paper ↗
ChipNet: Budget-Aware Pruning with Heaviside Continuous Approximations
Arnav Chavan*, Rishabh Tiwari*, Udbhav Bamba*, Deepak K. Gupta

Experience

Amazon Lab126Apr 2025 — Present
Senior Applied Scientist · San Francisco, USA
Work on extreme hardware–software co-design, leading generative-AI capabilities on Amazon's custom silicon to enable on-device GenAI for Alexa+.
Nyun AIAug 2023 — Mar 2025
Co-founder & CTO · Bangalore, India · acquihired by Amazon Lab126
Built compression pipelines for foundation models and their edge deployment, leading a team delivering solutions for Fortune-5 enterprises.
Microsoft ResearchAug 2022 — Feb 2023
Research Intern · Bangalore, India
Large-scale HTML understanding, and the impact of model compression on fairness in LLMs (ACL'23). Advisors: Dr. Sunayana Sitaram, Suresh Parthasarathy.
Google DeepMindJan 2022 — Apr 2022
Student Researcher · Bangalore, India
Empirical study on simplicity bias in deep networks — methods to quantify and mitigate it. Advisor: Dr. Pradeep Shenoy.
MBZUAIAug 2021 — Dec 2021
Research Assistant · Abu Dhabi, UAE
Compressing vision transformers via structured sparsity in a unified framework (CVPR'22). Advisor: Prof. Zhiqiang Shen.
Transmute AI ResearchMay 2020 — May 2023
Co-founding Member, Senior Researcher · Tromsø, Norway
Co-founded a research lab, mentored 10+ students, and drove work on network compression, meta-learning, and efficient ML. Advisors: Dr. Deepak K. Gupta, Prof. Dilip K. Prasad.

Honors

Kaggle Competitions Master — one of the youngest Indian masters, at 19 (2020).
Winner, Amazon ML Challenge 2021 — first national edition, 3,000+ teams.
KVPY Fellowship (2016) — top 1,000 high-school science students in India.

Service

Workshop OrganizerResource-Adaptive Foundation Model Inference (AdaptFM), ICML 2026; and Resource-Efficient Deep Learning for Computer Vision, ICCV 2023.

Competition Organizer — designed and ran the Efficient LLM Challenge at AdaptFM @ ICML 2026 (176 teams), and authored the results writeup.

Reviewer — NeurIPS, ICML, CVPR, ICCV, Pattern Recognition Letters.

Open Source — Lead of Nyuntam, Nyun AI's model-optimization suite.

CV