AI and observability projects built around faster incident understanding.

Designed and deployed an incident RCA pipeline that combines Prometheus and ELK log data with an LLM summarization layer to correlate alerts and draft runbook recommendations for on-call engineers.

Impact: Contributed to a 35% MTTR reduction during production outages.

Engineered a TensorFlow Transformer pipeline over Prometheus time-series telemetry to detect latency spikes and infrastructure saturation before customer impact.

Impact: Tuned for 94% precision with iterative threshold calibration.

Explored running Gemma 4 fully offline in a home-lab environment, pairing it with a log streaming layer to summarize outages and draft RCA steps without data leaving the infrastructure.

Impact: Built as a proof-of-concept for privacy-first incident response.

Built and operated an active-active MySQL Galera setup across Azure and two on-premises datacenters, using virtually synchronous replication and failover paths for high availability and disaster recovery.

Impact: Sub-second synchronization and zero downtime during node failures.

Automated AWS infrastructure provisioning and configuration management with Terraform and Ansible, standardizing repeatable deployments across platform environments.

Impact: Reduced provisioning time by 70% while improving deployment consistency.

Optimized Apache Airflow DAG scheduling and dependency management across high-volume production data pipelines.

Impact: Improved workflow success rates from 88% to 96% across 500+ daily pipelines.