📚 Engineering Playbooks & References
Complete production engineering playbooks containing deep architectural diagrams, constraints handling, and personal engineering notes out of production deployments.
Copyright & IP Notice The architecture, diagrams, and written content provided on this website and in the downloadable playbooks are the original intellectual property of Ambuj Kumar Tripathi. You may read, reference, and learn from these materials. However, reproducing, republishing, or claiming this architecture/content as your own work—without explicit written permission and proper attribution—is strictly prohibited.
Engineering Portfolio Disclaimer I don't claim to be a professor, nor am I pretending to be an 'industry visionary'. These aren't theoretical tutorials or polished bootcamp projects. They are simply my raw, field-tested engineering notes provided "as is" from building production RAG systems under strict constraints (512MB RAM, $0 budget). Use these insights at your own discretion as I do not guarantee their suitability for every production environment.
Building Real AI Systems
60 Pages • Architecture Playbook
The complete production playbook covering architecture, failures, and fixes across the entire RAG pipeline from ingestion to tracing.
📖 Read 60-Page PlaybookMaster RAG Engineering
11 Pages • Technical Reference
Personal engineering reference notes detailing document loaders, chunking strategies, embedding models comparison, and OOM prevention.
📘 Read 11-Page ReferenceAgentic Financial Parser LLD
45 Chunks • Technical Specification
The complete technical architecture, security layers, and 8-node LangGraph pipeline design for the Financial RAG system.
Agentic Financial Parser — Premium
Premium Edition • 2026
The premium engineering playbook covering the complete Agentic Financial Parser system — architecture, deployment, security, and production optimizations.
Indian Legal LLM — QLoRA Fine-Tune
14 Pages • Technical Documentation
Complete cell-by-cell walkthrough of fine-tuning Meta's Llama 3.2 1B using QLoRA on 14,543 Indian Legal QA pairs — trained on Google Colab free tier at ₹0 cost. Published on Hugging Face.
