Engineered Datadog tracing for real-time performance anomaly detection across deployed server infrastructure; automated alerts reduced customer service complaints by ~40%.

Built a churn prediction model on subscription and game usage data; triggered automated outreach to at-risk customers, reducing first-month churn by ~30% while maintaining profitability.

Built RAG pipeline for topic-wise assessment using citation gating; maintained 82% correctness on generated multiple-choice questions, driving 50%+ usage by all users during peak test periods.

Increased matchable retargeting audience by 62% by syncing app users to HubSpot Contacts and building a Streamlit dashboard for self-serve audience insights.

Formulated analytics REST APIs and data schemas for cohort reporting and risk segmentation; optimized queries via indexing and caching to cut analytics latency by ~35% under high concurrency.

Redesigned the Cohort Risk dashboard surfacing student engagement signals (login frequency, assignment progress) as visible instructor actions, driving 16% higher instructor dashboard engagement.

Consolidated 12 legacy PII endpoints to 2 authorized services via granular IAM access controls, reducing blast radius by 83% and achieving ISO 27001 readiness for the startup.

Engineered a RAG workflow agent that cut compliance-change turnaround time by 80% by detecting government portal updates and drafting citation-backed internal alerts using LangGraph.

Led 5-person team to predict SPE intensity for astronaut risk mitigation using LightGBM vs. TCN model comparison with time-series cross-validation.

Built end-to-end modeling pipeline: feature engineering → training → validation, selecting final model via time-series CV and error analysis.