Blog
Production engineering writeups from pTeachTech — RAG, agents, MLOps, AWS, BFSI compliance. The kind of detail you'd find in an engineering post-mortem, not a marketing blog.
Starts publishing June 2026 — alongside our first webinar.
Topics in the pipeline
Drafts in progress. Subscribers get the heads-up when each one lands.
The chunking-vs-truncation bug most RAG teams ship without noticing
Why your chunks need to be calibrated to the embedding model's token window, not the LLM's context window. The silent failure mode.
When to fine-tune — and when to absolutely not
A cost-benefit framework. The 80% of cases where RAG + better prompts beat $50K of fine-tuning. The 20% where it doesn't.
Hybrid retrieval that beats vector-only — without LangChain
BM25 + vector + RRF in 200 lines of Python. The retrieval pattern Azure AI Search uses internally.
BFSI AI compliance: what RBI / SEBI / DPDP actually require
Stripping the legalese. What needs to be in your AI deployment architecture if you ship for an Indian bank.
Eval-on-PR — making AI deploys regression-safe
A GitHub Actions workflow that runs Ragas on your golden dataset, fails the PR if faithfulness drops > 5%.
The agent failure-mode playbook
Loops, hallucinated tools, cost spirals — and the cheap guardrails that prevent them.
Get notified when posts go live
We use the same email list for webinars and new blog posts — one monthly email, not a daily newsletter. Subscribe via the webinars page.