Uday Lunawat

Led the transformation of a fragmented GenAI ecosystem into a production‑grade agentic AI platform for a Fortune 5 U.S. telecom.• Architected and scaled a LangGraph‑based multi‑agent orchestration layer enabling stateful, hierarchical reasoning.• Designed and implemented RBAC and secure secret‑isolation models, eliminating static credentials, enabling safe team agent sharing.• Built dynamic tool and runtime configuration injection, fully decoupling agent logic from environment‑specific secrets.• Unified backend services, agents, SDKs, RAG pipelines, and MCP servers under a unified modular Jenkins CI/CD pipeline.• Automated Cloud Run provisioning and secure access workflows, cutting environment setup time from days to 9̃0 seconds.• Enabled 13,000+ automated deployments and onboarded 300+ production agents within six months.• Leveraged Generative AI tools (Copilot/LLMs) to accelerate boilerplate code generation and automate tasks.

Led the implementation of a Human in the Loop system to enhance machine learning model accuracy through continuous human feedback integration, improving model performance and reliability., Managed and deployed annotation and training pipelines using Label Studio, FastAPI, Docker, and collaborative tools such as Miro board to streamline the machine learning workflow and enhance team collaboration., Implemented comprehensive CICD image scanning solutions using Trivy, seamlessly integrating with Google Cloud Platform (GCP) and Microsoft Azure, to identify and mitigate vulnerabilities, ensuring robust security and compliance in the continuous integration and deployment process.

Led the implementation of a Human in the Loop system to enhance machine learning model accuracy through continuous human feedback integration, improving model performance and reliability.
Managed and deployed annotation and training pipelines using Label Studio, FastAPI, Docker, and collaborative tools such as Miro board to streamline the machine learning workflow and enhance team collaboration.
Implemented comprehensive CICD image scanning solutions using Trivy, seamlessly integrating with Google Cloud Platform (GCP) and Microsoft Azure, to identify and mitigate vulnerabilities, ensuring robust security and compliance in the continuous integration and deployment process.

Developed a real-time liveliness service with features like face detection, recognition, matching, achieving low latency (150 ms), and leveraging AIOps pipelines for efficient deployment.
Enhanced MLOps with annotation tools, model registry, monitoring, training, and other best practices implementation.
Streamlined the development process by transitioning from individual notebooks to modularized scripts, facilitating better version control, collaboration, autodocumentation and integration into the production environment.
Developed and proposed the MVP Sept 2020 - Dec 2020.

Developed a real-time liveliness service with features like face detection, recognition, matching, achieving low latency (150 ms), and leveraging AIOps pipelines for efficient deployment., Enhanced MLOps with annotation tools, model registry, monitoring, training, and other best practices implementation, Streamlined the development process by transitioning from individual notebooks to modularized scripts, facilitating better version control, collaboration, autodocumentation and integration into the production environment., Developed and proposed the MVP Sept 2020 - Dec 2020.

Implemented a 24/7 automated In Process Quality Control (IPQC) Audit system utilizing an internal cloud service, along with back-end Python scripts. Effectively managed data and generated reports, significantly improving the efficiency of quality control processes., Designed Tableau dashboards for a Smart Factory initiative, providing real-time visualization of manufacturing data across 20+ vendor sites and forecasting using Machine Learning. Implemented monitoring alerts for key metrics, resulting in a remarkable 60x productivity boost for vendors and employees. This initiative saved approximately 200 hours weekly across all sites., Introduced a web-based file comparison tool to support quality control, streamlining processes and enhancing efficiency.

Implemented a 24/7 automated In Process Quality Control (IPQC) Audit system utilizing an internal cloud service, along with back-end Python scripts. Effectively managed data and generated reports, significantly improving the efficiency of quality control processes.
Designed Tableau dashboards for a Smart Factory initiative, providing real-time visualization of manufacturing data across 20+ vendor sites and forecasting using Machine Learning. Implemented monitoring alerts for key metrics, resulting in a remarkable 60x productivity boost for vendors and employees. This initiative saved approximately 200 hours weekly across all sites.
Introduced a web-based file comparison tool to support quality control, streamlining processes and enhancing efficiency.

Collaborated closely with developers and researchers, contributing with people all around the globe spanning 12 countries.
Utilized human activity recognition techniques to enhance user experience in interactive applications, leveraging machine learning models to detect and classify human actions.
Worked with multiple TTS frameworks such as NVIDIA Nemo, Google Wavenet etc. for Deepfake voice cloning for video game character.
Federated learning methodologies for collaborative model training, ensuring privacy and scalability in multiple simulated devices.