Developer II, Software Engineering
CurrentUST (formerly Abzooba) · Pune, India
- Designed and built an MCP (Model Context Protocol) server and client for modular AI tool use, and integrated Goose over streamable HTTP for real-time communication.
- Built the xpresso.ai MLOps platform in standardized Python, integrating Kubeflow, Airflow and GitLab, deployed on AWS.
- Containerized the platform with Docker and Kubernetes (deployments, services, ConfigMaps, PVCs) and automated build and deploy stages with Jenkins CI/CD.
- Split the system into microservices to cut latency and improve fault tolerance and availability, and added multi-model deployment for modular ML workflows.
- Shipped a zero-code ML component that lets users build models without writing code, speeding up development by 90%.
- Migrated a legacy Flask app to async FastAPI, improving performance and response time by 20%.
- Wrote a VS Code extension in TypeScript for working with the Xpresso platform from the editor.
- Delivered client ML solutions on Google Vertex AI and AWS SageMaker, and built a dynamic Dockerfile generator for Jenkins agents with selectable JDK and Python versions.