LLMs

Prompting Is Architecture: Why Understanding Systems Matters More Than Ever

A Spark story about scale, grain, and why speed makes system awareness non-negotiable.Image generated using ChatGPTRecently, I ran into a Spark job that looked perfectly fine. I used Claude in this case to generate code, which wasSyntactically correctSemantically correctEasy to read and reason aboutThis worked on the sample dataset. Assuming Spark will manage the scale, we pushed it to production. However, on prod, it took more than 3 hours to complete on a 40-node EMR cluster. It was trying to...

Read

Agentic AI 101

Agentic AI 101 explored the evolution of AI from traditional rule-based systems to dynamic, goal-driven agents powered by large language models (LLMs).The session introduced the foundations of Agentic AI, covering the seven types of AI agents and their real-world applications.Emerging Agentic Protocols and trends in agent interoperability were discussed, along with practical challenges like hallucinations, complexity, orchestration, and governance.Strategies for building resilient, trustworthy, and ethically aligned agentic systems were shared by the speakers, Mahita D and Swetha A from Sahaj...

Read

Democratizing Large Model Training on Smaller GPUs with FSDP

Preethi Srinivasan, a Solution Consultant at Sahaj Software, demonstrated how combining QLoRA and Fully Sharded Data Parallel (FSDP) enabled large model training on smaller, consumer-grade GPUs, making advanced AI research more accessible.The talk covered the evolution of parallelism strategies, from traditional data, model, and pipeline parallelism to advanced techniques like FSDP, which significantly reduced memory usage, training time, and communication overhead while improving GPU utilization.Preethi discussed quantization, mixed precision, CPU offloading, and activation checkpointing, highlighting the trade-offs between speed, accuracy,...

Read