Thinking in Tokens

Enterprises are racing to adopt GenAI, but most are still stuck in pilot purgatory. In this episode, we sit down with Sujeet — ML Architect at Dell and former Data Scientist at HP — to unpack what it actually takes to build a scalable GenAI stack for 2026.
We dig into the realities behind RAG architectures, NL-to-SQL systems, workflow orchestration, embeddings, time-series foundations, and serving LLMs as production-grade APIs. Sujeet breaks down the difference between flashy demos and platforms that ship, drawing on nearly a decade of building ML systems that impact global operations.

If you're building AI inside a large org, leading an ML team, or trying to take a GenAI initiative from POC to platform, this episode is your blueprint. No hype — just the architecture, tradeoffs, and lessons from the trenches.

Follow-
Bupender: linkedin.com/in/bhupender-sharma-02497723
Sujeet: linkedin.com/in/sujeet-jog-83276a5a
Codebase: https://www.codebase.com/

What is Thinking in Tokens?

Thinking in Tokens is a monthly podcast that explores the rapidly evolving world of artificial intelligence, from cutting-edge research to real-world enterprise applications. Hosted by Bhupender Sharma, Head of AI at Codebase.com, LLC, the show brings together deep technical insights and practical perspectives on how AI is transforming industries.

With extensive experience building large language models (LLMs) for global enterprises like Dell, Bhupender dives into the opportunities, challenges, and future of AI—one token at a time. Whether you’re a developer, business leader, or AI enthusiast, Thinking in Tokens offers thought-provoking conversations and expert analysis designed to keep you ahead in the age of intelligent machines.