Projects
Selected Work
The projects I put the most into — real architecture decisions, not just tutorials.
Cronbell (formerly PulseCheck) — Cron Job & Scheduled Task Monitoring
A dead-man's-switch monitoring service for cron jobs and scheduled tasks: each job pings a unique URL when it finishes, and Cronbell emails, Slacks, or webhooks you when a ping goes missing. Live as a free demo on Fly.io with Postgres on Neon, backed by a 52-test suite in CI.
- Dropped the always-on Celery worker for a token-guarded internal endpoint on a 15-minute external trigger — and found GitHub’s scheduler too unreliable to be the only one, so cron-job.org is primary and GitHub Actions the backup
- Web machine auto-stops when idle and Postgres sits on Neon, so both can sleep between checks; the 15-minute cadence is chosen for exactly that
- Slack, Discord, and webhook alerts, opt-in public status pages, and a per-account Prometheus /metrics endpoint for existing Grafana boards
- Rate limiting and CSRF protection on every state-changing endpoint, with a 52-test PyTest suite running in CI
FastAPIPostgreSQLSQLAlchemyRedisNeonGitHub ActionsDockerFly.io
NyotaNow — Invitations as a link, with RSVPs
A consumer product for Indian families: instead of forwarding an invitation JPG on WhatsApp, the host shares one link with a live countdown, Maps directions, and one-tap RSVP. Seven occasions (Navratri included) and seven card designs, with Hindi and English both first-class and real Devanagari typography. Live at nyotanow.in on Vercel, Postgres in Singapore.
- One card component renders the live preview, the invite page, and a 1080×1350 PNG export, sized entirely in container-query units so all three stay identical
- No accounts: each invite is guarded by a random key stored only as a SHA-256 hash, compared in constant time, with identical 404s for wrong and missing keys
- Hardened after launch: rate-limited public write actions, Sentry errors surfaced on a private stats page, real-user Speed Insights, and a Vitest suite gated by CI
- Dates and limits enforced server-side rather than trusted to the browser; guest data is deleted three months after the event while the card stays
Next.js 16React 19TypeScriptPostgreSQLDrizzle ORMVitestSentryVercelNeon
Azure Databricks Delta Lake Pipeline
Built a cloud-native data pipeline on Azure using Databricks, Delta Lake, and PySpark to ingest and transform large-scale log data with transactional and historical querying support.
- Implemented micro-batch processing for analytical workloads
- Designed a production-style log analytics workflow
- Used Delta Lake for versioned and reliable data processing
AzureDatabricksDelta LakePySparkSQL
Log Management Microservice
Designed a distributed log ingestion platform capable of processing thousands of asynchronous events with Kafka-driven streaming, real-time alerting, and analytics-ready storage.
- Built asynchronous pipelines for high-throughput ingestion
- Enabled decoupled producer-consumer architecture
- Created a scalable foundation for operational monitoring
DjangoKafkaMongoDBDockerMicroservices
Event-Driven Notification System
Built an async notification engine supporting email and SMS workflows through decoupled microservices and resilient message queue orchestration.
- Designed resilient message processing flows
- Improved reliability through decoupled services
- Focused on scalable workflow orchestration
DjangoKafkaMongoDBDockerAsync Processing
Capstone Todo API
A production-ready REST API for task management built with Go and PostgreSQL, containerized with Docker, and deployed on Render.
- Secure CRUD operations
- Cloud deployment readiness
- Clean architecture and API structure
GoDockerPostgreSQLRenderREST API