ThoughtsonSystems

Writing about AI architecture, deployment challenges, and lessons learned from the trenches of software engineering.

07 / BLOG

Insights&Writing

How I Used 418 Million AI Tokens for Just $0.64 — And How You Can Use Powerful Coding Models Almost Free

I used around 418 million DeepSeek V4 Flash tokens between August 5 and August 31, and my reported usage cost was only $0.64 through the $1/month Command Code Go Plan. In this guide, I’ll show you how to set up Command Code with LongCat 2.0 and OpenCode with Muse Spark 1.3 Contributor Free on Windows, macOS, and Linux—and use them directly inside VS Code.

2026-09-05

Machine Learning12 min read

Building a Probabilistic Football Prediction ML Engine

How I built an end-to-end football prediction ML engine using Poisson Regression, XGBoost, time-aware feature engineering, exact-score probability modeling, formation prediction, and starting XI analysis.

July 19, 2026

AI Engineering11 min read

How I Built an AI Photobooth That Scales to 30,000 Users on a Single Server

The model is the easy part. Here is the full system architecture — async job queues, bounded worker pools, load shedding and multi-key rotation — that let a Google Gemini photobooth survive 30,000 people uploading selfies at a live football event.

Jun 2026

AI Engineering8 min read

Building Production-Ready RAG Pipelines with LangChain and Pinecone

RAG systems are only as good as their retrieval architecture. Here's how I build pipelines that stay accurate at scale...

Jan 2025

NLP6 min read

BanglaBERT Fine-Tuning: Lessons from Building for Low-Resource Languages

Fine-tuning transformer models for Bangla taught me more about NLP fundamentals than any course. Here's what worked...

Nov 2024

MLOps5 min read

Why MLOps Is the Missing Piece in Most AI Projects

Most AI projects fail not because of bad models, but because of terrible deployment practices. Here's the infrastructure checklist I use...

Sep 2024

AI Engineering8 min read

How I Built a Production RAG Pipeline That Actually Works at Scale

Most RAG tutorials show you the happy path. Here's what I learned building a retrieval-augmented generation system that serves thousands of queries daily — including the parts nobody talks about.

2026-05-31

Case Study5 min read

How I Built a Complete Luxury Ride-Hailing Platform Solo — TriState Ride

I designed, built, and shipped a full Uber-like luxury transportation platform — passenger app, driver app, and super admin panel — entirely solo. Here's the full technical breakdown of TriState Ride.

2025-05-19

Read All Articles