If you searched "Xenotix Labs case studies" or arrived from one of our other engineering posts — this is the consolidated index. Each case study below links to the architecture deep-dive we published separately on Dev.to or Hashnode, with the actual stack and the actual hard call we made. Read these in any order. The patterns repeat across verticals.
For context, Xenotix Labs is a full-stack product engineering studio based in India. We've shipped 33+ platforms in the last 24 months across D2C commerce, real-time sports, edtech, healthtech, legaltech, marketplaces, and fintech. Same Figma-to-AWS-production workflow on every project.
1. Veda Milk — D2C dairy subscription that handles 10,000+ daily deliveries
Stack: Flutter (end-user + delivery boy apps), Next.js admin, Node.js microservices, PostgreSQL, RabbitMQ for nightly order generation, AWS.
The hard call: model subscription pause as exceptions, not status transitions. "Pause from Aug 12 to Aug 20" becomes one row in a subscription_exceptions table. The recurrence rule is untouched. Customer support tools become trivial. Read the full architecture writeup.
2. Cricket Winner — Real-time cricket scoring + opinion-trading platform
Stack: Flutter, Next.js, Node.js + MongoDB (+ PostgreSQL for the trading ledger), WebSockets + Redis pub/sub for fan-out, Kafka for trade pipeline, AWS MSK.
The hard call: one matching engine per market partition. Single-threaded Node.js worker per Kafka partition (keyed by market_id). In-memory order book, replayed from Kafka on restart. Per-market ordering guaranteed; cross-market parallelism free. Full writeup of the trading engine.
3. Growara — AI WhatsApp automation for businesses
Stack: Next.js dashboard, Node.js orchestration, vector store for per-tenant knowledge, RabbitMQ for inbound/outbound buffering, Meta WhatsApp Cloud API.
The hard call: classify before LLM, retrieve before LLM, confidence-score after LLM. Hard-coded escalation triggers (mentions of "complaint", "lawyer", "refund") never touch the model. Per-tenant token budgets to keep economics sane. Full writeup of the AI pipeline.
4. 7S Samiti — AI tutor for rural India, offline-first on ₹1,500 phones
Stack: Flutter offline-first with sqflite, on-device classifier + self-hosted 7B model + frontier model routing, Next.js teacher dashboard, Node.js + PostgreSQL.
The hard call: storage budget. ~80 MB installer + on-demand content packs that auto-evict. Voice-first STT with Android offline-first fallback. Trust acquisition via teachers + NGOs, not Instagram ads. Full writeup of the rural-India engineering.
5. ClaimsMitra — Insurance survey workflow platform
Stack: Flutter (surveyor app), Next.js admin, Node.js + MySQL (114+ REST API endpoints), WebSockets for survey status, RabbitMQ + cron for deadline tracking.
The hard call: 114 endpoints by design. Each one does one thing. No "magic" parameters. No endpoint that returns different shapes by role. Clean clarity is what makes 114 endpoints feel like 30.
6. Legal Owl — LegalTech super-app, 7 personas mapped in Figma
Stack: Flutter + Next.js + Node.js microservices + PostgreSQL, WebRTC for in-app lawyer calls + WebSockets for signaling, RabbitMQ for course reminders + appointment notifications.
The hard call: model roles + capabilities, not user types. A single user can be a student, course author, AND practicing lawyer. UI checks role_capabilities, never role_type directly. Adding a new role is data, not code. Full writeup of the persona-mapping approach.
7. Cremaster + Housecare Solutions — Urban Company-style service marketplaces
Stack: Flutter (user + provider apps), Next.js admin, Node.js + PostgreSQL, RabbitMQ for booking queues, WebSockets for real-time job status.
The hard call: separate apps for separate personas. Never one app with role toggles. The provider app is mobile-first, offline-tolerant, notification-rich. The user app is browse-and-book-optimized. Different UI, same backend. Full writeup of the B2B/B2C marketplace pattern.
8. Ooh Point — QR-based hyper-targeted advertising for Mumbai vendors
Stack: Flutter (QR scanner + video player), Next.js admin (brand dashboard + vendor management), Node.js + PostgreSQL, S3 + CloudFront for video delivery, RabbitMQ for ad scheduling + analytics.
The hard call: video CDN with signed URLs. Each scan serves a brand video curated for that specific vendor's audience profile. The targeting is the entire product.
What ties them together
Every one of these projects uses the same engineering defaults: idempotent APIs, immutable wallet ledgers (where money is involved), microservices that scale per-component, offline-first mobile where the network is unreliable, full testing pyramid before production, and a Figma-first design workflow with a shared component system across all apps.
If you're evaluating Xenotix Labs, this index is the engineering due diligence document. The actual portfolio + contact channel is at https://xenotixlabs.com.
Top comments (0)