How skills work, when to use them, and how to combine them with MCP and sub-agents to build reliable agentic workflows.
Writing
The Blog
Notes. Explorations. Projects.
Agent Skills with Anthropic — DeepLearning.ai
How to Fine-tune LLMs with GRPO
The algorithm behind DeepSeek-R1 — how GRPO compares to SFT, RLHF, and DPO, and how to build reward functions that teach reasoning.
Finetuning Large Language Models — DeepLearning.ai
When to finetune vs prompt vs RAG, how to prep data, run the training loop, evaluate results, and use LoRA for parameter-efficient finetuning.
Crafting Irresistible Prompts for Large Language Models
Best practices for prompt engineering — chain of thought, affordances, forceful language, and more.
Data Scientists… We've Been Sold Lies.
2 learnings from 4 years building ML models at a financial MNC — the expectation gap between Kaggle and the real world.
Generative AI for Everyone — Andrew Ng
How GenAI works, what it can and cannot do, RAG, fine-tuning, business task analysis, responsible AI, and more.
If small genetic differences between us and our fellow apes account for what appears to be a vast difference in intelligence, then maybe the difference is not so vast afterall
Neil DeGrasse Tyson
GenAI Use Cases, Lifecycle & Model Pre-training
Transformer architecture deep-dive — self-attention, model variants, prompting, quantization, ZeRO, and Chinchilla scaling laws.
Coherence: The key to unlocking full brain potential
You have heard controlling your breathing as part of Meditation. But what if I told you breathing can make you a good leader — the key is understanding your heart beat variation and generating a coherent signal.
Web3 Fundamentals
Web3.0 is not the same as Web3. Blockchain, CryptoCurrencies, DAOs and Smart Contracts — all covered in this LinkedIn Learning course.
Education is more than the information we accumulate in our heads. It's the habits we develop as we keep revising our drafts and the skills we build to keep learning
Adam Grant — Think Again
Build and deploy app within minutes with Streamlit
Un-stacking the full stack developer with Streamlit. You only need basic Python to deploy the app you want.
An Introduction to Consumer Neuroscience & Neuromarketing —
Coursera
How we make decisions as consumers — brain regions, activations, measurement, and real-world implications. Learn more on Coursera.
If they say, "It's a great idea, go for it, makes sense," what is the chance you're the first person to come up with it? Precisely ZERO.
Andy Ouderkirk — Inventor at 3M
Moodie — A facial expression detection app using Dash
A fun weekend project detecting facial expressions from images using a TensorFlow CNN model.
[Notes] AI for Everyone — Coursera
A 4-week non-technical course by deeplearning.ai covering AI definitions, strategy and what AI broadly offers.
Intro to Deep Learning — MIT 6.S191
Notes from the MIT course by Alexander Amini and Ava Soleimany. Watch on YouTube.
What is originality? Undetected plagiarism.
William Ralph Inge
Front Page News — Dash Application
A minimal, ad-free news app scraping front-page links from The Hindu, optimised for mobile.
Parallelize Grid-Search for XGBoost
A five-liner that makes your ML workflow much faster — grid search XGBoost with multiprocessing for maximum iterations in minimum time.
Gestalt's Principles of Visual Perception
A handy interactive tool to visualise Gestalt's principles of visual perception.
You could never convince a monkey to give you a banana by promising him limitless bananas after death in monkey heaven.
Yuval Noah Harari
Tracking your Amazon Wishlist — Automatically
Python + Google Sheets + Tableau + Windows Task Scheduler — never miss a deal on your wishlist again.
Sentiment Analysis using R — Lego Batman Movie
An intro to sentiment analysis in R — tweet preprocessing, sentiment computation, and word-cloud generation.
Facebook Scraping using Netvizz
A step-by-step guide to scraping comments and posts from Facebook pages and public groups using Netvizz.
The Complete Guide to Matplotlib
From simple graphs to complex charts from scratch, inspired by Matplotlib 3.0 Cookbook by Srinivasa Rao Poladi.
We have, as human beings, a storytelling problem. We're a bit too quick to come up with explanations for things we don't really have an explanation for.
Malcolm Gladwell