TALSON THOMAS

Implemented MLFLow to track the model versioning and KServ to serve the model.
Deployed and managed various data engineering tools to clean and process raw data ( Airflow, Trino, Hive-metastore, Spark, Flink, Querybook, arrow, Jupyterhub and kafka )
PostgreSQL Used for storing all experiment metadata, ensuring reliable and structured data storage.
MinIO Utilized for storing large binary objects such as trained models and other artifacts, providing scalable and efficient storage.
Deployed Keycloak OpenID - Provides a robust authentication mechanism to protect access to the tracking server.
Designed and implemented CI/CD pipeline to auto build and deployment to various kubernetes cluster environments.
Used Terraform and Ansible configuration management tools to manage cloud and native infrastructures. ( AWS and On-prem Environments )
Rancher Kubernetes management platform that simplifies the deployment and operation of Kubernetes clusters.
Harbour docker registry, highly available and scalable cloud native registry.
Implemented centralized package and library management using Pypi and Conda Registry, streamlining the efficient handling of dependencies in the software development lifecycle
Implemented RBAC and Pod Security Policies for access control and enhanced security.
Configured ELK stack (Elasticsearch, Kibana, Logstash) for effective log management.
Automated backup procedures for PostgreSQL databases and conducted Kubernetes ETCD backup and restore operations.
Implemented Grafana and Prometheus for monitoring, creating multiple dashboards and alerts.
Kong API Gateway to route requests and secure communications from clients to microservices.
Used Helm and Kustomize tools to manage kubernetes application deployments.