Insights, learnings, and analysis from my journey in data science, AI and software engineering.
Published · October 4, 2026
Every data scientist knows the phrase "garbage in, garbage out", and must have faced the issue at some point in their career. A sophisticated model trained on poor data will still produce poor predictions. RAG systems have the same problem, but it shows up in a different place.
Published · Aptil 25, 2026
Before the 2019 ICC ODI World Cup the thought appeared in my mind if ML can accurately predict the outcome of matches even before it starts. This was not just about guessing a winner but about testing a deeper hypothesis. Can the vast sea of historical sports data actually reveal the hidden variables that decide a match?
Published · Aptil 03, 2026
"A problem well stated is a problem half solved." — Charles Kettering
More often than not, issues in software do not arise because the code is wrong. They come from incomplete thinking behind the code. When the problem is understood clearly, the solution almost always becomes simpler.
Published · March 28, 2026
What Fred Brooks knew that machine learning is still learning
Back in 1975, Fred Brooks, in his seminal work "The Mythical Man-Month" wrote - dedicate half the schedule for "testing". His was not a suggestion, but a rule of thumb from decades of hard experience building real software.
“It’s harder to read code than to write it.” — Joel Spolsky
Published · July 1, 2022
Criclytics is a predictive analytics system built for cricket. It is the engine powering the predictive layer of cricket.com, a flagship product of Head Digital Works Pvt. Ltd.
A gist of what I learnt over the years from executing many projects in data science — practical takeaways that go beyond algorithms.
A mathematical analysis of batsmen performances in the Caribbean Premier League — grouping players by their playing style and effectiveness using clustering techniques.
Published · December 31, 2019
"I dream of painting and then I paint my dream." Vincent Van Gogh
A practical tutorial on using callbacks to control training in TensorFlow — stopping early when the desired accuracy is reached.