I own the backend at Guestara — a hospitality SaaS startup where hotel guests check in via smart locks and reservations sync in real time across property management systems. The kind of system where a 500ms timeout isn't a performance issue; it's a guest locked out of their room at 2 AM.
My stack: JavaScript/TypeScript · Node.js · MongoDB/PostgreSQL · Redis · AWS Lambda / SQS / API Gateway. I design BFF layers with Zod-validated contracts, wire IoT devices (Tuya smart locks, Seam), integrate third-party PMS platforms (Mews) via webhook pipelines, and ship it all serverless — in production, not in demos.
Outside work: 200+ consecutive days on LeetCode, going deep on distributed systems, and building production-grade portfolio infrastructure (RAG pipelines, multi-tenant SaaS backends) to push beyond what the sprint backlog allows.
API Design Versioned REST contracts with idempotency keys, cursor-based pagination, and typed error envelopes that tell callers exactly what failed and how to recover. An API is a promise to every caller — I write ones I can keep.
Data Layer PostgreSQL as the source of truth. Redis for the hot path. Query plans I actually read before shipping. Indexes aren't decoration — they're the gap between 12ms and 1200ms under load.
Reliability Engineering SLOs defined before dashboards are built. Distributed tracing before debugging sessions. Post-incident write-ups that make the next incident shorter. Observable systems aren't a luxury — they're the job.
Async & Scale Stateless horizontal services, SQS queues for async workloads, Lambda for the long tail, Redis pub/sub for real-time fan-out. Design goal: the queue absorbs spikes so the user never notices.
Systems Thinking Currently deep in distributed systems — consensus algorithms, consistent hashing, partition tolerance. Not just for interviews — for building things that don't fall over in ways you didn't predict.
Backend & Languages
Data Layer
Infrastructure & Cloud
Frontend — when the full stack demands it
Not just a streak — a discipline. 200+ consecutive days on LeetCode because consistent, deliberate practice is how intuition about hard problems is built. Not for the badge. For the muscle.
| Level | Topics |
|---|---|
| Advanced | Dynamic Programming · Divide & Conquer · Quickselect |
| Intermediate | Hash Table · Math · Database (SQL) |
| Primary Languages | C++ · TypeScript · PostgreSQL |
SupportIQ — Multi-tenant customer support automation built on RAG pipelines. Stack: LangChain, Anthropic API, pgvector, PostgreSQL, React. Real LLM infrastructure: semantic retrieval, conversation memory, tenant isolation, and a streaming chat interface. Not a tutorial rewrite — a system designed from the data model up.
→ Distributed systems internals: Raft consensus, consistent hashing, partition tolerance
→ LLM infrastructure: RAG pipelines, vector databases, Anthropic API, LangChain
→ Observability: OpenTelemetry, structured logging, distributed tracing at scale
→ System design: How Stripe, Uber, and Notion built what they built
"Reliability is a feature you ship on day one — not a patch after the incident."
"The API that surprises the caller has already broken its contract."
"A system is only as observable as you needed it to be at 3 AM."
"Correctness first. Performance second. Never the reverse."
If you're building systems where correctness, latency, and ownership actually matter — let's talk.
