At KPMG, I built an incident lifecycle module for an event-driven alert pipeline handling roughly 500 monitoring alerts daily across three systems. I added retry and dead-letter queues, alert flapping detection, deduplication, Jira ticket escalation, and MS Teams notifications to replace manual triage.
I also designed a five-agent natural-language-to-SQL system using Google ADK and Gemini 2.5 Pro, enabling non-technical users to query BigQuery in plain English. I contributed to a RAG-based alert classification layer using LangChain, FAISS, and a locally hosted Qwen 3 4B model.
At Sensearly.AI, I was the sole engineer building Python services to collect continuous patient telemetry from the Fitbit Web API, including OAuth 2.0 token refresh.
