My Journey

Replaced an existing Slack based error alerting system with Sentry, enabling centralized error capturing across our backends, visibility into the number of affected users, and clear prioritization of which errors to fix, cutting time to resolution from a couple of days down to a couple of hours.

Pioneered and led the migration from Lambda based APIs to ECS based Django, eliminating cold start latency and reducing response times from as high as 7 seconds down to roughly 300ms, with further improvements in some cases.

Engineered a new analytics pipeline replacing a Lambda job that scanned the entire database, using DynamoDB Streams into Firehose and S3 instead, reducing load on production tables and enabling larger scale analytical workloads.

Architected a shared notification platform decoupling backend services from the frontend across multiple channels (push, inbox, and event based background updates), using SNS/SQS fan-out so any upstream service could notify users without direct coupling to delivery logic, with dedicated per-consumer queues, filter policies, and retry handling to isolate failures and ensure reliable, timely delivery.

Developed the core infrastructure for securely exposing AI agents, with no prior team experience in the space, requiring rapid self-upskilling; built CI/CD pipelines, APIs for AgentCore Runtime access, and Gateways exposing tools for agents to use.

Implemented secure, JWT authenticated chat and session management ensuring strict isolation between users' agent sessions, enabling personalized outputs such as actionable plans generated from a user's genetic and lifestyle data.

Designed a CloudFront and S3 based asset distribution system with edge caching, serving app assets to users worldwide in milliseconds and reducing the app install/download size from 200MB to 45MB.