PinnedDevavrat Samak·Feb 12Don’t Ask Your Model to Learn Boundaries, but Build Them Into Your SystemA systems-level comparison of Mixture-of-Experts and routed small language models for regulated inference.
InTowards AIbyDevavrat Samak·May 24I Ran the Same Algorithm Ten Times. The Results Were All Over the Place.Part 2 of 2. If you have not read Part 1, it covers the KV cache problem and how TurboQuant works. This part is about what happened when I…
InTowards AIbyDevavrat Samak·May 19The Paper That Made Me Stop and Actually Think: Understanding TurboQuant and the KV Cache ProblemPart 1 of 2. The foundations. Part 2 covers what actually happened when I ran the experiments.
InTowards AIbyDevavrat Samak·May 12, 2025AI Meets Antiquity: Fine-Tuning Gemma-3 on Heritage Texts with Unsloth & Lightning AITL;DR: In this article, I share how I fine-tuned Google’s Gemma-3 on Peshwa-era Marathi texts using Unsloth and Lightning AI. You’ll learn…
Devavrat Samak·Mar 26, 2025Gemma2 Goes Multilingual: Fine-Tuning for English & Marathi TranslationsA] Introduction:-
Devavrat Samak·Jul 1, 2024Professor Claude 3.5 Sonnet..This project demonstrates the utilization of CrewAI agents, specifically a customized PDFSearchTool, with Anthropic's Claude 3.5 Sonnet…
Devavrat Samak·May 14, 2024RAG on Complex PDF using LlamaParse, Langchain and Llama 3 using GroqRetrieval-Augmented Generation (RAG):-
Devavrat Samak·Feb 23, 2024Demystifying Time Series Analysis: A Beginner’s JourneyHello folks! Thank you for visiting my medium blog. After a long time, thought of writing something….A response icon1A response icon1
Devavrat Samak·Jun 12, 2023My ML Journey….It all began with my close friend Ashwin Rachha, who introduced me to the fascinating world of Machine Learning. He patiently explained the…