Sakthivel N E

GenAI systems built at Virtusa now serve 500+ financial queries a day and cut analyst report time from 4 hours to 30 minutes. In 2.5 years I have shipped production RAG pipelines (LangGraph + Amazon Bedrock), lead scoring engines on 200K+ records, and multi-agent LLM architectures — each owned end to end. Strong in real-time ML inference, AWS cloud deployments, and measurable business impact. AWS ML Engineer Associate Certified (2026).

Associate Engineer - Virtusa Consulting Services

(2024-01)

• GenAI Financial Assistant (LangGraph + Amazon Bedrock): Architected a production LLM assistant using Graph-based RAG (LangGraph + Amazon Bedrock), improving query response relevance by 35% over baseline vector search.
• From 4 hours to 30 minutes: Designed Orchestrator + Memory agent for multi-step reasoning over financial documents — cutting analyst report turnaround
• Delivered sub-second real-time inference via Spring Boot APIs serving 500+ financial queries/day and reduced manual data-analysis effort by 45%.
• Lead Propensity ML System: Built and deployed lead scoring model (Scikit-learn + TabNet) on 200K+ records — improving sales prioritisation efficiency by 25% and achieving ROC-AUC of 0.82.
• Applied SMOTE, K-Means clustering, and pseudo-labeling on imbalanced data, improving model recall by 15%.
• Reduced model retraining cycle time by 25% through a modular, automated feature-engineering pipeline.
• Cloud Migration — IVR System (AWS Amazon Connect): Migrated legacy IVR workflows to Amazon Connect, reducing system downtime by 20% and lifting production uptime SLA to 99.5% across 3 client environments.
• Led end-to-end testing and validation across all client environments, achieving zero-defect production cutover and eliminating on-prem bottlenecks.

B.E. - Electrical & Electronics Engineering - R.M.K Engineering College (2023)