AI & Agentic Systems
0-to-1 build · evaluation · autonomous operations
- Transformed legacy, human-centric support workflows into an autonomous, AI-first operation, deploying an agentic system for multi-step triage, investigation and escalation that grew service margin.Meta
- Pioneered a production-grade (0-to-1) AI evaluation system to continuously measure, test and self-heal the LLM agentic skills used across the infrastructure support ecosystem — now also a benchmarking tool for LLM coding quality.Meta
- Built ML pipelines and MLOps lifecycle orchestration on AWS SageMaker — ML graph, explainable AI, fraud detection, privacy protection — including feature engineering and training-data labeling with client data scientists.EPAM
- Stood up a Generative AI competency center: drafted use cases and ROI metrics across product engineering and customer support, using unsupervised labeling models (zero-, one- and few-shot learning).Minerals