Dance studios needed a unified platform for student enrollment, scheduling, and payments. Existing solutions in the space were either prohibitively expensive or inflexible, forcing studios to rely on spreadsheets and manual reconciliation.

The system needed to support 100+ active enterprise clients, each processing $100K-$200K in monthly revenue, leading to a total platform processing volume of over $1M+ monthly. It required strict tenant isolation, a highly flexible custom discount engine, financial compliance, and operational clarity, directly competing against established players with a focus on extreme flexibility.

The platform was built as a multi-tenant SaaS system with strict data isolation and a clear separation of concerns.

Core backend: A NestJS-based monolith responsible for domain logic, orchestration, and tenant-aware APIs. NestJS was chosen for its strong modular structure, dependency injection, and long-term maintainability at this level of complexity.

Payments service: High-throughput payment workflows (Stripe webhooks, retries, reconciliation) were extracted into a dedicated Fastify microservice. This reduced coupling with the core backend, simplified testing, and allowed the payments pipeline to scale independently under load.

Frontend: A React application built around TanStack Query, enabling optimistic updates, cache consistency, and resilient UI behavior during partial failures.

Data layer: MongoDB is used as the primary datastore, with tenant-scoped access patterns and indexing strategies to enforce isolation and performance. While workable, the domain's relational nature surfaced clear limitations that informed later architectural discussions.

Equestrian facilities needed a sophisticated scheduling system for lessons, horse management, and instructor assignments. Complex domain logic involving recurring schedules, cancellations, waitlists, and resource allocation. Needed to handle timezone complexities and real-time availability updates.

Built with Hono for lightweight, fast API responses. React frontend with TanStack Query for real-time state management. Complex scheduling logic implemented with proper domain modeling. Real-time updates via WebSockets for availability changes. PostgreSQL with timezone-aware datetime handling.

Pointo operates large fleets of lithium-ion batteries powering e-rickshaws. Each battery continuously emits telemetry via a Battery Management System (BMS), producing high-frequency heartbeat data that must be ingested, processed, and analyzed in near real time.

The system needed to handle 5,000+ active batteries concurrently, with each device emitting a message every 30 seconds. This generated over 14.4 million heartbeat messages daily. We needed to support multiple IoT device manufacturers, enable real-time monitoring, and remain reliable under intermittent connectivity.

The platform was designed as a stream-oriented IoT ingestion and processing pipeline, optimized for reliability and extreme data velocity.

Edge & ingestion layer: IoT devices communicate via TCP. Custom parsers were implemented for each device type to normalize vendor-specific payloads.

Processing pipeline & Messaging: RabbitMQ buffers and fans-out the massive influx of heartbeat events, smoothing traffic spikes and decoupling ingestion from processing.

Core backend: A NestJS service handles business logic, persistence, and APIs, computing battery health and alerting thresholds.

Data storage: We engineered PostgreSQL to handle the massive data ingestion, implementing advanced indexing and optimization strategies to ensure complex search queries remained performant despite the rapidly growing dataset.

Infrastructure: Deployed on AWS using EC2, RDS, ECR, S3, and SES, with robust CI/CD pipelines.

Pointo's core operations — sales, customer onboarding, collections, and inventory management — were spread across disconnected tools and manual workflows. This resulted in data silos, delayed decisions, and operational friction as the organization scaled.

The platform needed to provide a single source of truth across teams, track battery and charger inventory accurately, support customer onboarding with KYC verification and digital contracts, manage collections and payment status, integrate with battery manufacturer APIs for remote immobilization, and offer real-time operational visibility without increasing manual overhead.

The system was built as a centralized internal operations platform, optimized for clarity, speed of iteration, and reliability.

Backend: An Express.js API handles core application logic and integrations. The focus was on predictable request flows and straightforward debuggability for a fast-moving operational environment.

Frontend: A React-based dashboard using TanStack Query for data fetching and cache management, enabling responsive UIs and consistent state across complex workflows.

Database: MySQL serves as the primary datastore, with a unified schema spanning sales, collections, inventory, and customer records — enabling cross-functional reporting and analytics.

Integrations: KYC verification services for customer onboarding, digital contract signing for customer agreements, and manufacturer APIs for remote battery immobilization based on payment or compliance status.

Infrastructure: Deployed on AWS with standard monitoring and alerting to support business-critical operations.