Recurring events system
Zero reported failures after rebuild, previously the top source of user complaints and iOS team blockers.
Technologies used
TypeScript
Nest
NodeJS
OpenAI
AWS Lambda
AWS S3
AWS Cognito
AWS DynamoDB
Motivation
The founder used AI coding tools to move fast and validate the idea. But when the iOS team tried to build on top of the backend, the cracks appeared: business logic scattered across handlers, a recurring events system that broke on edge cases, and no error handling to speak of. The codebase had served its purpose as a prototype but could not carry a real product. Azati was brought in to stabilize it and take full engineering ownership going forward.
Main challenges
Challenge 01
AI-Generated Code with No Architectural Boundaries
The initial backend was written with heavy AI assistance and lacked consistent structure. Controllers mixed business logic with data access, modules had no clear separation of concerns, and the codebase had grown organically without a guiding architecture, making every new feature a risk.
#1
Challenge 02
Broken Recurring Events Logic
The scheduling system, core to the product, had been partially generated by AI and partially patched manually. It failed silently on custom recurrence patterns, timezone edge cases, and override scenarios, causing incorrect reminders and lost user tasks.
#2
Challenge 03
No Error Handling or Observability
Lambda functions failed without logging. DynamoDB errors surfaced as blank screens on mobile. There was no structured logging, no alerting, and no way to diagnose production issues without reproducing them locally, a critical gap for a live consumer product.
#3
Our approach
Codebase Audit & Risk Mapping
Before touching a line of code, we spent the first week mapping every module: identifying AI-generated patterns that posed stability or security risks, documenting what worked, and building a prioritized remediation plan aligned with the iOS team's roadmap.
Surgical Refactoring, No Big Bang Rewrite
We refactored incrementally, module by module, without stopping feature delivery. Each area was stabilized, given clear interfaces, and covered with tests before the next was touched. The iOS team never experienced a blocked sprint.
Recurring Events System, Rebuilt from First Principles
We designed and implemented a flexible recurring events engine from scratch, supporting custom schedules, timezone-aware processing, per-instance overrides, and filter-based querying. This became the most reliable part of the system.
Personalized Maintenance Planning Algorithm
We implemented a smart scheduling algorithm that factors in season, local climate, pet ownership, household size, and usage patterns to generate balanced, personalized maintenance plans, distributing workload across the year without overwhelming users.
Observability & Error Handling Infrastructure
Centralized error handling was added across all Lambda functions. Structured logging, request tracing, and alerting were introduced so the team could diagnose and resolve production issues in minutes instead of hours.
Facing the same challenge?
Bring your complexity. We'll bring the plan. Select a convenient slot to start a conversation with our experts.
Schedule a callSolution
01
Flexible Recurring Events Engine
A purpose-built scheduling system that supports complex recurrence rules, per-event overrides, custom filters, and timezone-safe execution, replacing the fragile AI-generated logic that powered the original implementation.
Key capabilities:
- Custom recurrence patterns with override support
- Timezone-aware scheduling across user locations
- Filter-based event querying for mobile performance
02
Smart Maintenance Planning Algorithm
A personalization engine that builds maintenance schedules based on environmental and lifestyle signals: season, climate zone, pets, number of residents, and historical usage, distributing tasks evenly across the calendar.
Key capabilities:
- Multi-factor personalization logic
- Workload balancing across months and seasons
- Extensible rule system for future property types
03
Production-Grade AWS Serverless Backend
A refactored, fully owned NestJS backend running on AWS Lambda with proper module boundaries, typed interfaces, centralized error handling, and structured logging, built to be maintained and extended by any competent engineering team.
Key capabilities:
- Clean NestJS module architecture
- Centralized error handling and structured logging
- DynamoDB access patterns optimized for mobile query shapes
04
Ongoing Engineering Partnership
Azati continues as the backend engineering team, planning features, advising on architecture, and supporting system stability as the product scales. Full ownership, not a handoff.
Key capabilities:
- Active feature development and roadmap input
- Continuous system stability monitoring
- Direct collaboration with Product Manager and iOS leads
Results & business impact
Error visibility
Production issues now detected and diagnosed in under 15 minutes, compared to hours of manual reproduction before.
Feature velocity
iOS team unblocked, backend changes that previously took days to safely coordinate now ship in hours.
No rewrite required
Existing business logic preserved and stabilized, saving an estimated 3-4 months of redevelopment time.
Full engineering ownership
Azati owns the backend end-to-end, from architecture decisions to production incidents, with no dependency on the original AI-generated context.
