Writing on systems that have to hold.
I'm Siddik, a software engineer at Paygasus. I write about backend architecture, RAG and LLM pipelines, and machine learning in production - from real work, not tutorials. More about me
Claude Skills, Connectors, and Not Reaching for the Biggest Model Every Time
A working engineer's Claude Code setup: skills that kill repeated prompts, a token-compression plugin, routing small jobs to small models, and MCP connectors.
Read →
- 01
Designing Multi-Tenant Fintech Backends with Fastify and Cloud Spanner
Lessons from building a production multi-tenant fintech platform: tenant isolation strategies, RBAC in Fastify, Cloud Spanner data modeling, and CI/CD that pays for itself.
- 02
What Building a RAG Pipeline Taught Me About Search
Practical lessons from building DocuChat, a RAG-based PDF analysis tool with LangChain, Pinecone, and Next.js: chunking, metadata filtering, and cutting query latency 60%.
- 03
Squeezing 22% More Accuracy from E-commerce Recommendations with XDeepFM
How the Extreme Deep Factorization Machine (XDeepFM) improved e-commerce recommendation accuracy by 22% - feature interactions, CIN, and lessons from publishing in IJTE 2024.
- 04
Is MERN Still Relevant in the Age of AI!?
A fresher's honest take on learning the MERN stack in the AI era - why building things yourself still matters when models can scaffold an app in seconds.
- 05
Starting Your Coding Journey in 2024
Web development or data structures and algorithms? A practical guide for coding newcomers comparing Java full stack, MERN, and DSA paths - and how to choose.