Cloud Services & NestJS Experience

1. Which NestJS libraries and modules have you used, and what did you personally develop with NestJS?

Libraries: @nestjs/common, @nestjs/core, @nestjs/platform-express, @nestjs/typeorm (TypeORM + PostgreSQL), @nestjs/jwt, @nestjs/config, @nestjs/swagger, class-validator/class-transformer

I built multiple NestJS microservices covering domains like user profiles, calendar/scheduling, social interactions (follow/like), and search. Also developed a shared internal npm package for JWT authentication and role-based guards reused across services.

2. What technical and architectural decisions did you make in your NestJS projects?

Event-driven microservices via GCP Pub/Sub as inter-service message bus
Shared auth package — extracted JWT auth into a reusable npm package for consistency across all services
Timezone-aware scheduling — all dates in UTC, business logic in local calendar timezone
TypeORM datasource pattern — separate datasources for production (PostgreSQL) and tests (in-memory PostgreSQL via pg-mem)
Polymorphic role-based profiles — User entity with conditional relations to role-specific profiles
Pipeline architecture for search — chain of external providers → cache → deduplication → ranking
Event sourcing for data changes — every profile update publishes to Pub/Sub for cross-service cache invalidation

3. Which AWS, GCP, or Azure services have you used?

AWS: ECS Fargate, S3, SES, SNS, SQS, CloudWatch Logs, Secrets Manager, EC2

GCP: Cloud Run, Cloud SQL (PostgreSQL), Cloud Storage, Pub/Sub, Cloud Build, VPC, Load Balancer, MemoryStore (Redis)

Cloudflare: Workers, Browser Rendering (Puppeteer)

MongoDB: Atlas (managed), Atlas Search

4. Five or six cloud services and your responsibilities:

Service	Responsibility
AWS ECS Fargate	Designed task definitions for multiple service types (API, Queue, Admin, Public) across dev/stage/prod environments; built Prometheus service discovery that auto-detects running ECS tasks via ECS + EC2 APIs
GCP Pub/Sub	Implemented event-driven communication between microservices with typed payloads and multiple topics; responsible for reliability and message contract design
AWS S3	Built secure file upload flow: presigned URLs generated by backend, direct client-to-S3 upload; lifecycle policies and CORS configuration
Cloudflare Workers + Puppeteer	Built a serverless PDF generation service from scratch: HTML→PDF rendering, browser session pooling for 10x speedup, cron-based session pre-warming, aggressive resource blocking for performance
GCP Cloud Run + Cloud Build	Set up CI/CD pipeline: Git push → Cloud Build → Cloud Run deployment; Terraform infrastructure-as-code applied in the same pipeline; IAM service account management
MongoDB Atlas Search	Implemented full-text search with permission-filtered queries using Atlas Search indices

5. What monitoring and observability tools have you used?

Sentry — error tracking with profiling on both frontend (Next.js) and backend (Express)
Prometheus (prom-client) — custom metrics (Counters, Gauges) with tenant and organization labels; Node.js default metrics
CloudWatch Logs — custom Pino transport with batched writes and per-instance log streams
Pino — structured JSON logging with shared context propagation (request ID, tenant, organization, job, task)
Datadog — log aggregation and APM in Kubernetes environments. Configured structured Pino logs for Datadog agent ingestion; made an intentional design decision to omit hostname from log base fields in K8s so the Datadog agent correctly tags logs with the node host and preserves Kubernetes metadata (pod, namespace, container). Used Datadog for cross-service log correlation and debugging in production. Also integrated Datadog RUM (Real User Monitoring) and custom events on the frontend for tracking user-facing performance and business metrics.
Cloudflare Workers Observability — built-in metrics for edge workers

6. How involved were you in monitoring, troubleshooting, and improving the product?

I built the entire observability stack myself:

Wrote a custom Pino → CloudWatch transport
Built Prometheus service discovery that dynamically finds running ECS tasks and scrapes metrics from multiple services
Implemented contextual logging so every log entry carries tenant + organization for request tracing across services
Designed structured JSON log format compatible with Datadog agent auto-parsing in Kubernetes — ensuring logs are properly tagged with pod, namespace, and container metadata
Created health-check cron jobs for application heartbeat monitoring
Integrated Sentry with profiling for performance bottleneck detection
Set up log stream naming per-instance for easy troubleshooting

7. Did you work on the product continuously after the initial delivery?

Yes. The codebase shows sustained post-launch evolution:

50+ data migrations — date standardization to UTC, logo data migration, deal stage restructuring, permissions migration, cache invalidation, attachment extraction
40+ cron jobs added after launch — auto-enrich, contact synchronization, relation scoring, inbound email processing, scheduled report generation
PDF generation migrated from server-side rendering to Cloudflare Workers as a performance optimization
MCP (Model Context Protocol) integration retrofitted into an existing application for AI agent capabilities

8. What improvements, features, or optimizations did you implement?

Major features:

Recurring appointment scheduling with timezone-aware overlap detection
Multi-filter search with composite queries (JSONB fields, price ranges, full-text)
Full Microsoft Graph integration — Outlook/Exchange sync (emails, calendar, contacts) with deduplication
OpenAI integration — email information extraction, FAQ generation, chat completions
MCP server for AI agent interaction

Performance optimizations:

PDF generation moved to edge workers with session pooling — 10x reduction in generation time
MongoDB Atlas Search indices for full-text queries
TypeORM composite queries with aggregation for complex search results

Infrastructure & DevOps:

Full observability stack from scratch (Pino + CloudWatch + Prometheus + Sentry + Datadog)
Prometheus ECS service discovery with automated task detection
Migration framework with CLI runner and per-migration test coverage
CI/CD pipeline: Git → Cloud Build → Cloud Run + Terraform apply
Infrastructure as Code for all cloud resources (GCP via Terraform)