Senal Weerasinghe Email & Phone Number

Sri Lanka

Sri Lanka

H2O MLOps and Monitoring product developmentThis product has main features of deploying machine learning models to the production environments and monitoring their performance (model health, drift, etc) with the feature of Model Monitoring.Key Responsibilities1. Implemented a monitoring back-end service using SpringBoot and gRPC.2. Developed a Transaction Transmitter using the AKKA actor framework.3. Implemented REST APIs using Django/FAST API.4. Enhanced the performance of the Monitoring backend service through Asynchronous data processing.5. Contributed to front-end development of Monitoring Front-end using React.6. Contributed to MLOps front-end development using the H2O.ai Wave™ SDK.7. Maintained 80% unit test coverage for owned components using Junit, Mockito, and Pytest.8. Improved release efficiency by implementing CI/CD automation with Jenkins.9. Automated cloud deployment using terraform/Helm.10. Implemented monitoring for EKS cluster resources with AWS CloudWatch.11. Conducted critical issue debugging and analysis. (E.g. network disconnections and packet losses)Arogya AI Backend Implementation 1. Inventory Manager and Order Manager are AI application which were designed for hospital chains and used Apache Airflow to automate its workflows. 2. Automated entire workflow from training the models and storing predictions in Apache Redis using Apache Airflow. So a user just has to upload their dataset to S3 and the Airflow workflow will be triggered and final predictions are available in Redis to used by the front-ends.Technology StackOpen-source applications: Spring Boot, Apache Airflow, gRPC, SQLAlchemyDatabases: Apache Redis, Postgres, InfluxDB, Apache DruidLibraries: Python locust for performance automation Programming Languages: Python 3.7, Java 17CICD: JenkinsAWS: S3

Millennium Data Platform (MDP)Millennium Data Platform is a set of open-source applications which are used in realtime data streaming, processing and storing in the Millennium Surveillance application. Key responsibilities1. Implement high throughput Kudu Data Producer with exactly one delivery guarantee. 2. Performance tuning Kudu and Impala cluster to handle real-time data which is 150K messages per second. 3. Implement Java API called "SGrunner" to integrate open-source third party application with in-house built monitoring and orchestration framework called Millennium SysGuard. 4. CI/CD automations on Jenkins with terraform. 5. Design and implement cucumber based BDD testing framework to test high availability and fault ingestion for MDP applications. Millennium Surveillance 5.1 Data Warehouse Replacement project.This project was done to replace 15+ years old, in-house built data warehouse stack due to the scalable and security limitations. Key Responsibilities1. Implement fault-tolerant Apache Kudu data ingestion library. (KucuClientWrapper) 2. Cutover data ingestion components in Millennium Survellance product to ingest data into Kudu using "KuduClientWrapper" library.3. Performance test automation using TestNG and NMon.4. Cloud security - Our dataset were classified as highly restricted. Therefore, Authentication(Kerberos) and encryption was enabled for each and every channels used to communicate.Technology StackDatabase: Apache Kudu, Apache ImpalaAPI: Kudu client C++ API, Kerberos, Apache RangerProgramming Languages: C++, Java, PythonCICD: JenkinsAWS: Oracle/Postgres RDS, NLB, EFS with provisioning, S3

Sri Lanka

Market Health View (MHV)Market Health View is a real-time market monitoring dashboard with two-dimensional multi-lined graphs to provide a birds-eye view of the market in terms of Order and Trade statistics to compare market trends.Key responsibilities1. Implement high throughput Kafka Data Producer with exactly one delivery guarantee using "librdkafka" library. 2. Performance tuning of Kafka cluster to hand real-time data which is 150k messages per second. 3. Implement standard security mechanisms for Kafka, Zookeeper, and Spark. (SSL, SASL_SSL, SASL_SCRAM).4. CI/CD automation on Atlassian Bamboo.5. Design and implement cucumber based BDD testing framework to test high availability and fault ingestion in MHV components.6. Implement a python based stat analysing tool to record hardware utilisation (CPU, Memory, Disk) during the performance tests and publish them on a Grafana dashboard. 7. Evaluate hardware footprint of MHV system on production rates and process distribution was done based on the evaluation. 8. Define and finalise NFR(Non-functional requirements) of the MHV (Market Health-View) product.Technology StackMessage broker: Apache Kafka,Structured Streaming: Apache SparkDatabase: Hive, PostgresSQLProgramming Languages: Java, Scala, Python, C++CICD: Atlassian Bamboo

Sri Lanka

1. Performance test automation.2. Developed Java based system statistics analysing tool.3. Analysing memory, CPU of each thread of backend process and plotting graphs of each stats to ease up issue analysing.

Bit, Information Technology

Bachelor Degree in Information Technology.

High School, Biology/Biological Sciences, General