ML Blog

Welcome to my machine learning blog! Here you’ll find in-depth articles on various ML topics, from fundamental concepts to advanced techniques.

Table of Contents

• 1. Climbing the Ladder of Abstraction: My Computationally Irreducible Journey
• 2. A Deep Dive into ML Model Assumptions
• 3. Inductive Biases of ML Models
• 4. Designing a Low-Code ML Platform: Lessons from Building EcoKI

Article Details

Climbing the Ladder of Abstraction: My Computationally Irreducible Journey

Published: August 2025 | Reading Time: 15-20 minutes

A deep dive into the four paradigms of scientific discovery, from empirical experimentation to computational simulations and data-driven AI. Explore how my journey mirrors the evolution of scientific discovery and the importance of computational irreducibility.

A Deep Dive into ML Model Assumptions

Published: August 2025 | Reading Time: 15-20 minutes

A comprehensive guide exploring the fundamental assumptions behind major machine learning model classes. Learn about linear models, tree-based models, time series models, and more. Understand what happens when assumptions break down and how to address violations.

Inductive Biases of ML Models

Published: August 2025 | Reading Time: 10-15 minutes

Explore how inductive biases shape machine learning models and influence their learning capabilities. Understand the trade-offs between different architectural choices and how they affect model performance.

Designing a Low-Code ML Platform: Lessons from Building EcoKI

Published: December 2025 | Reading Time: 12-15 minutes

Lessons from designing a low-code ML platform for industrial IoT. Explore the five-layer architecture, key design patterns (Strategy, Composition over Inheritance), and the trade-offs involved in building production-grade systems that balance power with accessibility.

Featured Resource: LLM & RAG Systems

Note

🔗 Advanced Retrieval and Re-ranking

A comprehensive documentation site covering the latest research and practical techniques for building RAG (Retrieval-Augmented Generation) systems. Topics include:

• Two-Stage Pipeline Architecture — Retrieval for candidate selection, re-ranking for precision

• Building RAG Pipelines — From MVP to production-ready systems

• Hard Negative Mining — Advanced techniques for training dense retrievers

• Cross-Encoders & LLM Re-rankers — State-of-the-art re-ranking methods

• 30+ Library Comparisons — LangChain, LlamaIndex, ColBERT, and more

• Benchmarks & Datasets — BEIR, MS MARCO, Natural Questions

Upcoming Posts

• Reinforcement Learning Fundamentals - From Q-learning to policy gradients

• Production ML Systems - Deployment, monitoring, and maintenance strategies