Saikarthik Palagani

Deep Multi-Modal Microservice Trace Anomaly Detection Framework

Designed a multi-modal deep learning framework combining structural graph topology and latency dynamics for microservice trace anomaly detection. • Built hierarchical span feature extractors integrating call-graph tree depth, service-operation tokens, and non-linear duration normalization. • Implemented self-supervised pre-training and dynamic stratified threshold calibration to handle severe class imbalances across large-scale distributed systems. • Benchmarked against state-of-the-art models (TraceVAE, GTrace, PUTraceAD) across standardized industry datasets.

Scalable Multi-Vendor E-Commerce Backend

• Developed ascalable multi-vendor backend supporting Customer, Seller, and Admin workflows with role-based access control (RBAC). • Prevented inventory overselling using PostgreSQL row-level locking. • Improved API performance through Redis caching, cursor-based pagination, and database query optimization. • Implemented JWT authentication and Celery background workers.