Senior Backend Engineer

Department: Engineering

Experience: 4-5

Skills: AI & LLM Integration, react JS, AWS Cloud Infrastructure, B2B, MERN Stack, Marketing Automation, Claude, mongodb, Node.js (Express.js)

• Build and own APIs: Design, ship and version Express.js REST APIs for dashboards, POS integrations, campaigns and partner webhooks, with validation, auth, rate limits and clear contracts.
• Run event pipelines: Keep Kinesis Data Streams, Lambda consumers and SQS workers correct under load, with idempotent writes, retries, dead-letter queues, backpressure and replay.
• Choose the right store for each workload: Keep transactional data in MongoDB and move heavy analytics to ClickHouse (materialized views, AggregatingMergeTree, the right ORDER BY keys). Use S3 for archives and bulk exports.
• Own cost: Know what each service, query and job costs per month. Find the top spenders in AWS and ClickHouse Cloud, cut them and show the savings in numbers.
• Own performance: Profile slow endpoints and queries using MongoDB explain plans and ClickHouse query_log. Fix indexes and access patterns, and hold p95 latency targets.
• Plan for scale: Design for 10x data volume with sharding, partitioning, TTL and retention policies, hot/warm/cold storage tiers and autoscaling for ECS workers.
• Protect data correctness: Find and fix race conditions, double-counting and drift between MongoDB and ClickHouse, especially in money-like data such as loyalty points and credit wallets.
• Raise the bar: Write short design docs, review pull requests carefully, mentor engineers and help shape how the team uses AI coding tools.

• API: Node.js, Express.js, Joi/AJV, JWT/Passport, Swagger
• Transactional data: MongoDB with Mongoose (aggregations, indexes, transactions)
• Analytics: ClickHouse Cloud with the @clickhouse/client Node.js client (MergeTree family, materialized views, dictionaries)
• Streaming: AWS Kinesis Data Streams, Lambda consumers
• Background jobs: AWS SQS, ECS Fargate workers, EventBridge Scheduler, Redis + Bull
• Storage: AWS S3 (imports, exports, archives, presigned URLs)
• Quality: Jest, Supertest, ESLint, Prettier, Husky

• Deploy and infra: Serverless Framework, Docker, ECS autoscaling, SSM
• Observability: Datadog APM (dd-trace), CloudWatch, Winston

• 5+ years of backend engineering, including 3+ years running Node.js and Express.js services in production for real customers.
• Deep MongoDB expertise. You can read an explain plan, design compound indexes in the right field order, write aggregation pipelines that avoid full collection scans, and know when a document is too big or a collection needs sharding.
• Hands-on ClickHouse experience, or another columnar store you’ve run in production (BigQuery, Redshift, Druid, Snowflake). You understand sort keys, partitions, materialized views, and why FINAL and JOINs get expensive.
• Experience with AWS event-driven systems (Lambda, Kinesis Data Streams, SQS, S3). You know shard limits, batch sizes, visibility timeouts, partial batch failures, cold starts and dead-letter queues.
• Cost thinking. You’ve cut a real cloud bill and can explain how, whether through right-sizing, batching, caching, retention, storage tiering or moving work off the wrong database.
• Performance and scale experience. You’ve taken an endpoint or job from seconds to milliseconds, or from 100K to 100M rows, and measured it before and after.
• Strong API design skills: resource modelling, pagination, idempotency keys, versioning, consistent errors, and security basics (OWASP, input validation, NoSQL injection, rate limiting).
• Correctness under concurrency. You recognise lost updates, double-processing and race conditions, and fix them with atomic operators, transactions, idempotency or locks.
• Clear communication. You can write a one-page design doc covering trade-offs and costs, and explain a production incident without blame.

• You own every line. You read and understand each diff the agent writes before it goes to review.
• Evidence before “done”. Tests pass, queries are benchmarked on real data volumes and edge cases are checked. You don’t take the agent’s summary on trust.
• Safe by default. Agents get read-only access to production data. No secrets go in prompts, and destructive commands need a human.
• Context engineering. You keep CLAUDE.md, skills and plans current so the agent improves for the whole team.
• Token cost awareness. You choose the right model and effort for the task, just as you would right-size an EC2 instance.

• Change data capture from MongoDB to ClickHouse (PeerDB, ClickPipes, Debezium) or Kinesis Data Firehose into S3
• Data lifecycle work: TTLs, archiving to S3 or Parquet, and hot/warm/cold tiering with defined retention rules
• Building AI features on the Claude API or AWS Bedrock (tool use, prompt caching, structured output, vector retrieval)
• Infrastructure as code (Serverless Framework, CDK or CloudFormation) and ECS autoscaling policies
• Messaging at volume: WhatsApp Business API, SMS gateways, email delivery and webhooks
• Payments or billing systems (Razorpay, Chargebee), or any ledger-style data where every unit must reconcile
• Experience in B2B SaaS, retail tech, restaurant tech or loyalty products
• Open-source contributions, or Claude Code skills and plugins you’ve written and shared