Topic pillar · 3 posts

Building B2B SaaS in India — Engineering Lessons

Frank engineering lessons from shipping VIZIQO Assist and VIZIQO Campaign: stack selection, cloud infrastructure at each scale, RAG in production, and India-specific compliance overhead.

Building B2B SaaS in India means making engineering choices that look different from the typical Silicon Valley playbook. The market accepts simpler infrastructure (a single t3.xlarge can take you to ~200 concurrent users for under ₹15,000/month), but the compliance overhead is heavier — DLT registration alone takes 2–4 weeks, WhatsApp Business API approval has its own timeline, and Razorpay's edge cases around GST and RBI auto-debit will cost you two weeks if you don't plan ahead.

If your product is AI-powered, the production realities of RAG add another dimension: chunking strategy decides retrieval quality, embedding costs grow 10x faster than user count without caching, and hitting sub-2-second voice latency requires aggressive optimization (pre-warming, payload filtering, re-ranking, streaming).

This pillar collects our engineering writing — what we learned shipping two production SaaS products in 12 months, the infrastructure decisions at each growth stage, and the LangChain + Qdrant gotchas that no tutorial mentions.

Key takeaways
  • Spend one week on stack selection, not more — pick technologies your team knows well.
  • Stage your infrastructure: single EC2 → managed RDS + ALB → ECS Fargate → EKS. Don't skip stages.
  • Allocate ~15% of the first six months to compliance — DLT, WhatsApp Business API, data localization, GST.
  • In RAG, chunking strategy matters more than embedding model — spend 40% of dev time on chunking/preprocessing.
  • Never use a technology you can't debug at 3 AM when production is down.

In this topic

Engineering · 9 min read · Mar 14, 2026

Building a SaaS Product in India: What We Learned Shipping Two Products in 12 Months

From choosing the right stack to handling payments, compliance, and infrastructure — a frank account of what it actually takes to build and ship B2B SaaS in India. Lessons from building VIZIQO Assist and Campaign.

Engineering · 9 min read · Mar 10, 2026

LangChain + Qdrant in Production: What No Tutorial Tells You

We built VIZIQO Assist's knowledge base on LangChain and Qdrant. Here's what actually broke in production — chunking strategies, embedding costs, re-ranking, and the latency issues we didn't expect.

Engineering · 8 min read · Mar 3, 2026

Cloud Infrastructure for SaaS Startups: What to Use at Each Stage of Growth

From a single EC2 instance to Kubernetes and multi-region — the infrastructure decisions you make early will either scale with you or bottleneck you. A practical guide to cloud architecture for SaaS founders.