Vikas Chandra

Vikas Chandra leads AI at Atoms, with work focused on efficient language models, reasoning, and on-device AI. His writing explores how model training, compression, and inference strategies can make capable AI systems practical at smaller scales.

Career

Cosigned byLoading...

CosignedSign in to view

Elsewhere

Atoms

Sep 2026 — Present

Vice President of AI

Carnegie Mellon University

Ph.D. in Electrical and Computer Engineering

Products & media

v-chandra.github.io

Diffusion Models Learn to Think

An argument for testing diffusion language models as reasoning systems, with a method for making reinforcement learning training more practical.

Writing

v-chandra.github.io

Scaling Down Beats Scaling Up: The Algorithmic Attack on AI's Memory Wall

Explores algorithmic approaches to inference memory limits, including why efficiency improvements can matter alongside new hardware.

Writing

v-chandra.github.io

Does Reasoning Require Scale?

Examines how training methods and inference strategies can enable smaller language models to perform reasoning tasks.

Writing

v-chandra.github.io

The Personal Context Graph: Why On-Device AI Will Capture the Layer That Cloud Models Can't

Presents a perspective on how on-device AI could use personal context and why that opportunity may differ from enterprise applications.

Writing

v-chandra.github.io

The Audio-Visual Gap in Embodied AI

Discusses multimodal understanding of first-person video and data methods for connecting audio and visual context.

Writing

Profile name

Company name

Company name

Profile name

Company name

Company name

Profile name

Company name

Company name

Profile name

Company name

Company name

Profile name

Company name

Company name

Vikas Chandra

Atoms

Carnegie Mellon University

Products & media

Diffusion Models Learn to Think

Scaling Down Beats Scaling Up: The Algorithmic Attack on AI's Memory Wall

Does Reasoning Require Scale?

The Personal Context Graph: Why On-Device AI Will Capture the Layer That Cloud Models Can't

The Audio-Visual Gap in Embodied AI

Curated Lists

List Inclusions