Suman Debnath
Director, Developer Experience and Product at Crusoe
Suman Debnath is a Principal Machine Learning Advocate at Amazon Web Services, where he focuses on advancing Natural Language Processing (NLP), Large Language Models (LLMs), and Retrieval-Augmented Generation (RAG). He is passionate about leveraging open-source technologies such as TensorFlow, PyTorch, Numpy, LangChain, and Pandas to drive innovation in machine learning. With a strong presence in the global AI community, Suman has spoken at over 100 international events, including PyCon, PyData, ODSC, and meetups across multiple countries. Currently, he is part of AWS's Global AI Engineering Advocacy team.
Talks
Data Con LA 2026
The Physics of LLM Inference: From First Principles to Cluster Scale
Why latency spikes the moment concurrent users hit an endpoint, and how continuous batching, PagedAttention, and distributed serving keep throughput high, with live demos of Managed Inference and fine-tune-to-deploy workflows on Crusoe Cloud. We will also dive into Crusoe MemoryAlloy, which reinvents KV caching for cluster-scale inference.
Decentralized AI: Safer, Faster, and Greener
As AI systems move from centralized clouds into personal devices, edge compute, and distributed networks, a pivotal question follows: can decentralization make AI not just more powerful, but safer, faster, and greener? Builders of personal, self-hosted AI, climate-tech and product leaders, and sustainable AI infrastructure experts examine the concrete patterns, from personal AI servers and embodied agents that keep data local, to AI-powered climate platforms, to low-carbon GPU clouds powering sovereign AI stacks, and what it takes to align incentives, protocols, and infrastructure with privacy, resilience, and sustainability.
Data Con LA 2024
Building AI Solutions with Multimodal Retrieval-Augmented Generation (RAG): Unifying Text, Image, and Video Search
Discover how to implement multimodal Retrieval-Augmented Generation (RAG) using Amazon Bedrock to unify text, image, and video search. Learn practical AI solutions that solve real-world challenges like document analysis and image/video content retrieval.
Data Con LA 2022
Building a serverless data processing pipeline with PySpark on cloud