SA.
  • Home
  • About
  • Projects
  • Blog
  • Resources
  • Contact

Blog

Thoughts on machine learning, deep learning, and building ML systems in production.

Oct 25, 2025

How to Avoid Messing Up Feature Order in Neural Network Serving: Building Robust & Reliable Pipelines

If you've ever trained a neural network on tabular data, you've probably encountered this subtle yet dangerous problem: feature ordering...

Read on Medium →

Oct 23, 2025

When Neural Networks Take the Easy Way Out: How to Stop Your Model from Being Naive

Sometimes, the smartest thing a neural network does... is take a shortcut. Explores how neural networks often take the easiest path to minimize loss.

Read on Medium →

Research Publication

arXiv · Aug 16, 2021

Efficient Feature Representations for Cricket Data Analysis using Deep Learning based Multi-Modal Fusion Model

Investigated the use of adaptive embeddings to represent inter-related features for IPL cricket data analysis using a deep representation learning framework with contrastive loss.

View on arXiv →

Built by Souridas A