Playground

Linear Regression Primer

Linear regression is the simplest useful model in machine learning. It contains most of the ideas that show up everywhere else in ML: a loss function, an optimizer, regularization, and the bias-variance trade-off.

Cloudflare Deployment Modes for a CRUD App

Cloudflare gives you three real ways to ship a full-stack CRUD app: Pages with Functions, Pages with a standalone Worker, and Workers with static assets. They look similar but differ in routing and auth, in where your code lives and when it runs.

Class of One: Podcast feed from NotebookLM summaries

Podcasts are a good first pass to get oriented on a topic. I can convert some of my technical reading goals into podcasts then I can grok them better when I decide to go deeper into it.

Point-in-Time Joins

Intro

Jiggy: a desktop agent wearing a slime costume

There is a small blue blob living on my desktop. It floats above every window, including full-screen ones, on every virtual desktop. Left alone, it wanders, naps in a corner, or peeks off the edge of the screen and comes back. Move the cursor too close and it gets startled. Type for a while and it politely scoots out of the way of your caret. Pick it up and it bounces happily when you let go.

Move US Stocks from INDMoney to Interactive Brokers

I wanted to move my US stocks (securities) from INDMoney to Interactive Brokers. I didn’t find any clear steps documented anywhere. It took me multiple attempts to finally be able to move over my portfolio.

PyTorch Fundamentals - Week 5

Brushing up on my PyTorch skills every week. Starting from scratch. Not in a hurry. The goal is to follow along TorchLeet and go up to karpathy/nanoGPT or karpathy/nanochat. Previously,

PyTorch Fundamentals - Week 4

Brushing up on my PyTorch skills every week. Starting from scratch. Not in a hurry. The goal is to follow along TorchLeet and go up to karpathy/nanoGPT or karpathy/nanochat. Previously,

PyTorch Fundamentals - Week 1, 2, & 3

Brushing up on my PyTorch skills every week. Starting from scratch. Not in a hurry. The goal is to follow along TorchLeet and go up to karpathy/nanoGPT or karpathy/nanochat. Summary of the 1st three weeks.

Life Logging: Calls

I am building a comprehensive set of tools to do life logging. General idea is:

[Mini] Life Logging

I am building a comprehensive set of tools to do life logging. General idea is:

Consolidated Recommendation Systems

This post is a quick summary of Lessons Learnt From Consolidating ML Models in a Large Scale Recommendation System. I have also added a few questions I got while reading it. I end the post with what we do at work to deal with this.

Document Your Progress at Work

How can you ensure that your contributions are also recognized?

Confidence Intervals and Coverage

Confidence Interval (CI)

Lognormal to Normal Distribution

The Normal and lognormal distributions are fundamental concepts in statistics. I recently used the relationship between these two distributions in a project. In this blog post, I want to share what I learned.

Ooty: Friendships, Travel and Painting 📍

I spent the last two weeks of August in Ooty📍, A hill station in Tamil Nadu. Transitioning from Chennai’s heat to Ooty’s cold within a day was drastic. My hoodie was happy to be out from the bottom of my bag.

Visualizing a GroupBy (or a Bipartite Graph)

Have you ever needed to present the output of a GroupBy or Pivot Table?

[Mini] How to Parse JSON in Spark without Knowing the Schema?

Problem Statement

Fitness Dashboard with Google Fit

In this post, I will describe my Google Sheets dashboard, where I track all fitness-related aspects.

Mechanisms for Data Science Projects

Good intentions never work, you need good mechanisms to make anything happen - Jeff Bezos.

Belagavi or Belgaum 📍

Entering Belagavi My first interaction in the city: ₹250 for an auto for 2 KMs. That’s just robbery! He came down to ₹100 after I told him off. Then found another auto guy who said ₹80. (This was also high, but I was tired.) Belagavi is the (proposed) 2nd capital of Karnataka. It is a controversial city: both Maharashtra (Maha) and Karnataka want it within their borders. I don’t understand this…

[Summary] Monolith: Real-Time RecSys With Collisionless Embeddings

Paper link: Monolith: Real Time Recommendation System With Collisionless Embedding Table

Learning How to be a Mentor

Some people are born coaches. I am not one of them.

[Summary] Deep Recurrent Neural Networks for OYO Hotels Recommendation

Paper link: Deep Recurrent Neural Networks for OYO Hotels Recommendation

Passing Multiple Parameters in PySpark MapPartitions

Alternate title: k-Nearest Neighbours (kNN) in PySpark

Personalising the Swiggy Homepage Layout - Part I

Personalising the Swiggy Homepage Layout - Part II

Workation in Pondicherry 📍

📆 28th April

All the Small Things

I delivered this speech on Sunday (17th Apr) at a Toastmasters Meeting. The objective was to introduce body language and vocal variety in my delivery.

Toastmasters Speech with a Purpose

Today, I am going to discuss my second Toastmasters speech. The first speech is here.

Toastmasters Icebreaker Speech

Last year in November, I joined Toastmasters (TM) to build my communication and leadership skills. One part of the TM is the prepared speeches. You have to prepare and deliver a short speech in front of an audience. You also receive feedback from an evaluator.

Improving data science productivity with dslib

Git Basics

I created this introduction to Git for the Data Science team of Swiggy. I believe, the better we know Git, the more efficient our engineering processes will be. Data Scientists come from a variety of educational backgrounds. Many of them will not know anything about git. Many others will just have a working knowledge; they will use git commands, but they won’t really know what really happens.…

Minutes - Building the History Page

In the last two posts [1, 2], I had described the app I want for personal tracking and my progress on the app. I have to create five tabs: History, Dashboard, Tracker, People, and Settings. I have finished designing the Settings page. The Tracker page is in the making. Today, I am going to talk about the design of the History page.

Minutes - Building the Settings Page

In the previous post, I had talked about the Quantified Self and how I started working on an app for that. I had shown the bottom navigation with the following five screens on the homepage:

Minutes - A Quantified Self App for Myself

I’ve long since 2014 engaged in some form of Quantified Self. It started with a simple Libre Office worksheet. I tracked my finances, sleep, time spent on entertainment, walking steps, and many other things. Maintaining a worksheet had a lot of flaws. I didn’t want to put my data in a Google worksheet, so it always remained on my laptop. Sometimes, I forgot to enter the data, and other times, I…

Building and Hosting a Podcast Website for Free

A few months back, I was looking for some good Data Science podcasts, and the only sources I found were blog posts with reviews about a few podcasts. Many of those podcasts were no longer active. A few hosts explicitly mentioned that on their homepages, and some just stopped creating new episodes. Apple Podcasts, Spotify, Pocketcasts - where you can subscribe to a podcast - none of them showed if…

Lang. Identification in Code Mixed Text - Lit. Review 3

I am reviewing the literature available on language identification for multilingual documents, focusing on Indic languages. I’ll try to cover it in chronological order, but there might be a few misses here and there. After a decent coverage of the research, I expect to have enough understanding to discuss the challenges present in this task in a code-switched setting and its importance.

Lang. Identification in Code Mixed Text - Lit. Review 2

I am reviewing the literature available on language identification for multilingual documents, focusing on Indic languages. I’ll try to cover it in chronological order, but there might be a few misses here and there. After a decent coverage of the research, I expect to have enough understanding to discuss the challenges present in this task in a code-switched setting and its importance.

Lang. Identification in Code Mixed Text - Lit. Review 1

I am reviewing the literature available on language identification for multilingual documents, focusing on Indic languages. I’ll try to cover it in chronological order, but there might be a few misses here and there. After a decent coverage of the research, I expect to have enough understanding to discuss the challenges present in this task in a code-switched setting and its importance.

Design Thinking and Life

At the start of July 2020, I enrolled in a Coursera course on Design Thinking called Design Thinking for Innovation. Today, I’ve finally finished it. This post is a brief description of what I learned and my final submission about the reflection on one of the four design tools covered in the course that I employed in a challenge/problem of my choice.

Configuration Management in Python

There are a lot of ways to maintain configuration in Python project. I’ve recently learnt to do it in way that is not something new that I’ve found, but it was new to me. I’ll discuss my progression from basic hard-coding constants in a project to this new method.

Python Module Versioning and Releases

At work, these days, I am building some new Python ML modules to be used within other projects. When I am making modules for myself, I don’t care about the versioning or releases. But working in an environment when others are using your libraries then versioning is required. This is what I know now which I wish I had known before starting.

Flask in Production

After coding your flask app, when you do flask run, you get the following warning:

Packing a Backpack For Cycling Trips

This will be my first post about cycling. I started cycling almost a year back. I have done about a dozen of half-a-day trips. They usually range between 50-130 KMs, in and around Bangalore. This post is about what I have learned to carry with me on these trips. For some of the stuff there’s a back story as well. The final list is at the bottom.

Gradient Descent

Recently, I realised that I have used majority of classical machine learning algorithms using scikit, but I haven’t actually implemented them myself. I know the basics of the algorithms, but were I to implement them, I’d most likely fail. There are always a lot gotchas. So, I am going to start with it now. The first stop is Gradient Descent.

How to Design a Game With a Purpose

This post is the next in series of my literature study of the Human Computation and Games With a Purpose field. Following is the list of previous posts I wrote on this:

PyData Bangalore Meetup #7

PyData (Bangalore) is a community of users and developers of all things data - Data Science, Data Engineering, Machine Learning, Deep Learning, Data Ethics, Visualization, etc. We gather to discuss how best to apply Python tools, as well as those using R and Julia, to meet the evolving challenges in our use cases. We all get together every month with 4 talks, a few lightening talks to share ideas…

Human Computation and Games with a Purpose

I first read the terms Human Computation and Games with a Purpose while reading about crowd sourcing. Further reading on these terms and a discussion with my boss led me to Luis von Ahn, the inventor of reCAPTCHA and the founder of Duolingo. He is a pioneer in this field. I wanted to study more about this field of research and hence I begin my systematic study of it. Picking up Luis’s papers was…

Fixed Perimeter + Maximum Area = Square

I’d like to verify using Gradient Descent, that given a perimeter value of a quadrilateral, square is the one with the maximum area. This can be verified/proved using various analytical methods, but my objective here was to verify it using Gradient Descent. You ask why? Because I wanted to do it. In the process, I got more than what I had hoped for. Here are some intuitive explanations of the…