Atanu Maity·Jun 20Most “AI Agents” in Production Are Actually LLM-Powered Workflows — And That’s Perfectly FineOver the last year, “Agentic AI” has become one of the most overused terms in the AI ecosystem. Every other product seems to claim to be…
Atanu Maity·Apr 19Local Window Attention or LWALocal Window Attention, a variant of Attention mechanism in Transformer, has a lot of beautiful technical depth hiding under its simple…
Atanu Maity·Nov 9, 2025Why Many ML Hyperparameters Are Powers of 2?You often see dimensions like 32, 64, 128, 256, 512 in neural networks. This is not a coincidence. It comes directly from how computer…
Atanu Maity·Jun 28, 2025Geometric Interpretation of Feed-Forward Network in Transformer ArchitectureThe Expand and Contract (Intuition) of the Vector Space
Atanu Maity·Jun 21, 2025How LLMs achieve Near Human Level CognitionWhen people say Large Language Models (LLMs) like GPT-4 exhibit near-human cognition, they’re typically referring to their ability to…
Atanu Maity·Sep 22, 2024How Numpy vector operations are so fast and efficientHere we will consider one particular operation, which is ‘numpy.dot’, a core function in NumPy used for performing dot product operations…
Atanu Maity·Jun 2, 2024RAG Refinement using Re-RankerChatbots powered by Large Language Models (LLMs) can see significant improvements in their performance by incorporating external knowledge…
Atanu Maity·Dec 2, 2023Building A Data Driven Story: San-Francisco Crime ScenarioHere I have done a detail analysis on San Francisco crime dataset and have looked for some insights from the story hidden behind the data…
Atanu Maity·Dec 2, 2023Want Data from Youtube? Great Idea! Lets have itLike social media platforms, Youtube also has become a good interest for Data Science guys to collect data. By Data , I mean Number of…
