--- title: "Aakanksha Chowdhery: A lesser-known titan of the world of AI" description: "Highlights Aakanksha Chowdhery’s impact on Pathways, PaLM, and large-scale AI systems, showcasing the engineering behind today’s most powerful models." author: Ayushman Dash date: 2025-11-15 canonical: https://aiwithayushman.com/blog/aakanksha-chowdhery-a-lesser-known-titan-of-the-world-of-ai --- # Aakanksha Chowdhery: A lesser-known titan of the world of AI ## Why You Haven’t Heard Her Name Yet Chowdhery is not a household founder or brand face. According to her personal site, she was technical lead on the 540 billion-parameter model PaLM, and contributed to the infrastructure project Pathways at Google ([Google Research](https://www.achowdhery.com/)) Her role: designing, training, scaling, stabilising large models and supporting the unseen plumbing that makes them work. ## What She Built: The Infrastructure Behind The Buzz ### 1. Pathways: coordination at massive scale Pathways is Google’s “single system” vision to train a unified model capable of thousands of tasks and modalities. [A Silicon Valley Insider](https://asiliconvalleyinsider.com/2023/02/04/the-pathways-system-googles-nextgen-ml-platform-for-llms/) The paper behind PaLM describes how Pathways enabled a 6144-chip TPU v4 setup and achieved high hardware utilisation. [Google Research](https://research.google/blog/pathways-language-model-palm-scaling-to-540-billion-parameters-for-breakthrough-performance/) Chowdhery is listed among the authors for the Pathways system. [Journal of Machine Learning Research](https://www.jmlr.org/papers/volume24/22-1144/22-1144.pdf) ### 2. PaLM: from research paper to real capability The PaLM model doesn’t just represent size. It set benchmarks across reasoning, code generation, multilingual tasks, and few-shot learning. [Google Research](https://research.google/blog/pathways-language-model-palm-scaling-to-540-billion-parameters-for-breakthrough-performance/) Chowdhery co-authored the PaLM paper and, per sources, handled infrastructure/training strategy for it. [Journal of Machine Learning Research](https://www.jmlr.org/papers/volume24/22-1144/22-1144.pdf) ### 3. Multimodal systems & beyond Her profile lists work on PaLM-E (an embodied, multimodal model) and other cross-modality systems. [Aakanksha Chowdhery](https://www.achowdhery.com/) That means bringing together text, image, code, etc, not just language models in isolation. ## Why This Matters - **Infrastructure matters**: Most public AI discourse is about new algorithms or models. But without systems that can train, serve, cost-effectively scale, it remains research. She focused on the “boring but vital” work of engineering rather than just architecture. - **Scale unlocks new capabilities**: The PaLM paper shows that bigger isn’t just “more of the same”, performance on complex reasoning and code tasks jumped. [arXiv](https://arxiv.org/abs/2204.02311) - **Representation matters**: Engineering leadership in AI still skews towards a few names. Highlighting those who build the infrastructure widens the narrative. - **Product vs demo gap**: Many high-end demos don’t translate to real systems. She worked on making models that **run** not just look good. - **Transparent research ethic**: Google published detailed papers on Pathways/PaLM. This stands in contrast to some models where core specs weren’t disclosed. The difference isn’t trivial in responsible AI debates.

Her work is what it takes to serve a model to the planet. At the other end of that scale: I built 5 AI agents whose only job is to send my mother a good-morning message every morning at six. Check it out ↗ ## Key Takeaways for Professionals & Leaders - If you’re in AI or data leadership: ask not only *what* the model does, but *how* it’s trained, *where* it will run, *how* it scales. - For non-tech leaders: recognise that “data + model” is only part of the value. Infrastructure cost, serving latency, and reliability are equally strategic. - For anyone building or joining AI teams: there’s huge value in roles that focus on “make it real” (engineering, scaling, inference, tooling) rather than just “make it novel”. ## Final Word When you hear names like “GPT”, “Gemini”, “multimodal models”, remember there are people behind the scenes making them work at scale. Aakanksha Chowdhery’s work is a prime example of that. Not just in concept, but in systems that deliver. If you’d like, I can pull together a timeline of her major publications + talks, or analyse how Pathways/PaLM architecture informed later models. ## Discover More Insightful Reads Want to dive deeper into how leaders can harness AI strategically and effectively? Check out the blog at [**AI with Ayushman**](../blog) where you’ll find articles on: - [You're Viewing AI Wrong: Why It’s a Growth Engine, Not a Cost Cutter](./you-re-viewing-ai-wrong-why-it-s-a-growth-engine-not-a-cost-cutter) - [5 Prompts That Will Help You Crack Any AI Interview](./5-prompts-that-will-help-you-crack-any-ai-interview) - [Grokipedia: The AI Encyclopedia With Zero Accountability](./grokipedia-the-ai-encyclopedia-with-zero-accountability) Feel free to explore and bookmark the blog for regular updates meant for business leaders, AI strategists and decision-makers. ## Related reading - [Ashish Vaswani: The Lesser-Known Titan Who Built the Future of AI](/blog/ashish-vaswani-the-lesser-known-titan-who-built-the-future-of-ai) - [Four Nobel Prizes for AI in one year. Who got them and why?](/blog/four-nobel-prizes-for-ai-in-one-year-who-got-them-and-why) - [I worked on this 9 years ago and today it is used in the new Meta RayBan.](/blog/when-i-worked-on-this-9-years-ago-i-had-no-idea-that-it-will-be-used-in-the-new-meta-rayban.)