A semantic and lexical search engine for indexing and retrieving context-aware results from large volumes of structured and unstructured data. It supports complex queries with filters using vector similarity and lexical similarity, and can provide context-aware answers to questions based on search results. Data is ingested from sources like S3 buckets and websites. Then cleaned, chunked and optimized before creating embeddings for semantic search (RAG pipeline).

Key challenges included achieving indexing time of 1 million records in under 2 hours and delivering search results with AI-powered answers within 500ms. The solution involved trying out different transformers, distance types, and index types, as well as reranking results and providing configurable indexing options depending on the source data and customer requirements.

Key responsibilities:

• Led the design and implementation of data ingestion pipelines, document transformation processes, and indexing workflows.
• Developed semantic search APIs enabling flexible querying with metadata filters.
• Developed semantic, lexical search and AI answer features on top of GDN features.
• Contributed to and reviewed the React-based GUI console for configuring and testing indexing and search functionalities.

Languages: Python, React
Tools:

• Macrometa GDN – Data store, Vector store and underlying search provider
• FastAPI – HTTP framework and the interface to the users
• Docker & Kubernetes – For deployment and scaling, including the production environments
• GitHub Actions + Argo – CI and CD

An AI agent designed to automate web GUI testing and reduce repetitive QA work on frontend applications. Built as a Crew AI multi-agent system, it interprets feature requirements and QA scenarios, then generates Cypress tests. The process starts by providing clear test documentation and a live app URL to browser-use, which identifies the necessary actions. Another agent then converts these steps into a Cypress test suite.

Key responsibilities:

• Led and developed this tool as part of a broader framework for automating QA, collaborating with other team members on the overall system.

Languages: Python
Tools:

• Browser-Use – External API contract design and validation
• CrewAI - Multi agent framework

A scalable multi-tenant identity and access management framework for microservices, built to securely handle thousands of authentication requests with <10ms latency on dual-core, 256MB–512MB hardware. It featured a distributed apikey/token cache implemented using apache pulsar streams natively supported by the Macrometa GDN.

Key responsibilities:

• Led the architecture and development, focusing on tenant isolation, performance, and scalability.
• Designed and implemented the v2 of the framework, prioritizing Single-Sign-On (SSO), resiliency, and API clarity.
• Developed and improved API key/JWT flows, caching, observability, and metrics.
• Main contact point for auth framework issues on the platform.
• Contributed to React based SSO login components as needed.

Languages: Rust, React
Tools:

• Tokio – Rust async runtime
• Macrometa GDN – Tenant data modeling and storage
• Prometheus – Metrics collection
• Auth0 – Multi-option SSO provider
• Kubernetes – Containerized development and deployment
• GitHub Actions + Argo – CI/CD
• Apache Pulsar – Message queue/broker to keep local caches up-to-date

A cloud-native API gateway for microservices. Choreo Connect separates control and data planes for scalability, security, and performance. It supports dynamic configuration via gRPC-based xDS protocol, OpenAPI definitions, and integrates with WSO2 Choreo. Features include authentication, rate-limiting, observability, analytics, and distributed tracing. The gateway reliably handled ~3000 transactions per second with latency under 250ms for 1000 concurrent users on constrained hardware (1 CPU, 512MB memory).

Key responsibilities:

• One of the founding engineers
• Contributed to architecting the core product with scalability, performance, runtime mutability, and ease of maintenance as primary goals.
• Led the control plane development in Go.
• Contributed to the data plane in Java; filter chain implementation, authentication, and observability using Envoy ext_auth filter.

Languages: Go, Java, Protobuf
Tools:

• gRPC + xDS Protocol – to stream API and routing configurations from control plane to gateway in real time
• OpenAPI – for defining API structure and generating gateway configs
• Kubernetes – for deployment orchestration and service mesh integration
• Prometheus – for gateway telemetry and monitoring
• Jaeger – for distributed tracing
• OAuth2/JWT – for authentication and policy enforcement
• Envoy – as the base TCP proxy to implement gateway filter chain
• GitHub Actions – for CI/CD automation and test pipelines

A cloud-native, developer-centric API gateway for microservices that simplifies composing, securing, and monitoring microservice endpoints as managed APIs. It supports OpenAPI-driven configuration, runtime service discovery, legacy-to-modern API transformation, and can be deployed as lightweight, stateless containers in CI/CD pipelines.

Key responsibilities:

• One of the founding engineers
• Drove the architecture and initial implementation of the Microgateway core and toolkit.
• Defined the OpenAPI parsing engine, built the security and other API QoS features.
• Led containerization (Docker) and Kubernetes deployment strategies, established CI/CD workflows, and collaborated with UX and DevOps teams to expose a unified developer experience and rich observability.
• Took ownership of managing the open-source community around the project.

Languages: Java, JavaScript, Bash
Tools:

• OpenAPI Spec – for declarative API definitions and gateway config generation
• Docker – to package the gateway runtime as lightweight containers
• Kubernetes – for orchestrating and scaling gateway instances
• Prometheus, Jaeger, Zipkin – for collecting metrics and tracing
• JMeter – to load-test gateway performance under high throughput
• GitHub Actions – to automate build, test, and release pipelines

A fully featured API management solution including security, rate limiting, API analytics, and developer portals. The gateway component could handle 2500 TPS per instance while all other QoS features were turned on.

Key responsibilities:

• Contributed as a full-stack engineer throughout most aspects of the project (Specially around auth, rate limiting, subscription management, workflow management, publisher and developer portals).
• Visited multiple customers as an onsite product specialist.
• Mentored many juniors and interns.

Languages: Java, React
Tools:

• MySQL, PostgreSQL, MsSQL – as the data store for all API management components
• OpenAPI, WebSocket, GraphQL – support for all types of APIs and their governance
• JMeter – to load-test gateway performance under high throughput

A semantic and lexical search engine for indexing and retrieving context-aware results from large volumes of structured and unstructured data. It supports complex queries with filters using vector similarity and lexical similarity, and can provide context-aware answers to questions based on search results. Data is ingested from sources like S3 buckets and websites. Then cleaned, chunked and optimized before creating embeddings for semantic search (RAG pipeline).

Key challenges included achieving indexing time of 1 million records in under 2 hours and delivering search results with AI-powered answers within 500ms. The solution involved trying out different transformers, distance types, and index types, as well as reranking results and providing configurable indexing options depending on the source data and customer requirements.

Key responsibilities:

• Led the design and implementation of data ingestion pipelines, document transformation processes, and indexing workflows.
• Developed semantic search APIs enabling flexible querying with metadata filters.
• Developed semantic, lexical search and AI answer features on top of GDN features.
• Contributed to and reviewed the React-based GUI console for configuring and testing indexing and search functionalities.

Languages: Python, React
Tools:

• Macrometa GDN – Data store, Vector store and underlying search provider
• FastAPI – HTTP framework and the interface to the users
• Docker & Kubernetes – For deployment and scaling, including the production environments
• GitHub Actions + Argo – CI and CD

An ETL workflow framework on Macrometa GDN that automates data extraction, transformation, and loading between document collections and external sources. It supports log-based CDC and batch ingestion, schema inference, transformations using C8QL, and enables real-time data sync across Oracle, MySQL, PostgreSQL, MongoDB, Snowflake, and more.

Key responsibilities:

• Designed and implemented the workflow management and controller layer for Connection Workflows, including all external APIs for the controller that enable users to configure, trigger, and monitor data connector workflows.
• Contributed to the Meltano-based workflow execution and Argo Workflows orchestration layer.
• Contributed to or owned many of the available connectors.
• Contributed to GUI development and debugging as needed.

Languages: Java, Python
Tools:

• Macrometa GDN – storing data and executing transformations using stream workers
• Oracle, MySQL, MongoDB, PostgreSQL – syncing records with each database
• Meltano – ETL definition framework
• PyPI – hosting connector packages: https://pypi.org/search/?q=macrometa-&o=
• Argo Workflows – workflow orchestration engine
• React – implementing required GUI components for the GDN console

A Flutter-based cross-platform mobile application (iOS, Android) for browsing, viewing, and listening to Buddhist audio and video sermons. The app integrates with the existing backend supporting the website https://waharaka.com. Users can browse sermons by tags or keyword search. It includes a built-in audio player with seek, play, and pause functionality, as well as background playback. Users can also download audio sermons to their devices. Video listings are served via YouTube APIs, redirecting users to the YouTube app for playback. The app currently has over 5k downloads and a 4.9 rating from more than 270 reviews.

Key responsibilities:

• Responsible for approximately 90% of the application’s requirements gathering, design, and development.
• Currently working on version 2.0 in my free time, focusing on enhancing the user experience and adding new features.

Languages: Dart
Tools:

• Flutter – Cross-platform app development framework for Android and iOS
• YouTube APIs – To display live and uploaded video lists
• Android and iOS Developer Consoles – For publishing the applications

A scalable multi-tenant identity and access management framework for microservices, built to securely handle thousands of authentication requests with <10ms latency on dual-core, 256MB–512MB hardware. It featured a distributed apikey/token cache implemented using apache pulsar streams natively supported by the Macrometa GDN.

Key responsibilities:

• Led the architecture and development, focusing on tenant isolation, performance, and scalability.
• Designed and implemented the v2 of the framework, prioritizing Single-Sign-On (SSO), resiliency, and API clarity.
• Developed and improved API key/JWT flows, caching, observability, and metrics.
• Main contact point for auth framework issues on the platform.
• Contributed to React based SSO login components as needed.

Languages: Rust, React
Tools:

• Tokio – Rust async runtime
• Macrometa GDN – Tenant data modeling and storage
• Prometheus – Metrics collection
• Auth0 – Multi-option SSO provider
• Kubernetes – Containerized development and deployment
• GitHub Actions + Argo – CI/CD
• Apache Pulsar – Message queue/broker to keep local caches up-to-date

A semantic and lexical search engine for indexing and retrieving context-aware results from large volumes of structured and unstructured data. It supports complex queries with filters using vector similarity and lexical similarity, and can provide context-aware answers to questions based on search results. Data is ingested from sources like S3 buckets and websites. Then cleaned, chunked and optimized before creating embeddings for semantic search (RAG pipeline).

Key challenges included achieving indexing time of 1 million records in under 2 hours and delivering search results with AI-powered answers within 500ms. The solution involved trying out different transformers, distance types, and index types, as well as reranking results and providing configurable indexing options depending on the source data and customer requirements.

Key responsibilities:

• Led the design and implementation of data ingestion pipelines, document transformation processes, and indexing workflows.
• Developed semantic search APIs enabling flexible querying with metadata filters.
• Developed semantic, lexical search and AI answer features on top of GDN features.
• Contributed to and reviewed the React-based GUI console for configuring and testing indexing and search functionalities.

Languages: Python, React
Tools:

• Macrometa GDN – Data store, Vector store and underlying search provider
• FastAPI – HTTP framework and the interface to the users
• Docker & Kubernetes – For deployment and scaling, including the production environments
• GitHub Actions + Argo – CI and CD

A cloud-native API gateway for microservices. Choreo Connect separates control and data planes for scalability, security, and performance. It supports dynamic configuration via gRPC-based xDS protocol, OpenAPI definitions, and integrates with WSO2 Choreo. Features include authentication, rate-limiting, observability, analytics, and distributed tracing. The gateway reliably handled ~3000 transactions per second with latency under 250ms for 1000 concurrent users on constrained hardware (1 CPU, 512MB memory).

Key responsibilities:

• One of the founding engineers
• Contributed to architecting the core product with scalability, performance, runtime mutability, and ease of maintenance as primary goals.
• Led the control plane development in Go.
• Contributed to the data plane in Java; filter chain implementation, authentication, and observability using Envoy ext_auth filter.

Languages: Go, Java, Protobuf
Tools:

• gRPC + xDS Protocol – to stream API and routing configurations from control plane to gateway in real time
• OpenAPI – for defining API structure and generating gateway configs
• Kubernetes – for deployment orchestration and service mesh integration
• Prometheus – for gateway telemetry and monitoring
• Jaeger – for distributed tracing
• OAuth2/JWT – for authentication and policy enforcement
• Envoy – as the base TCP proxy to implement gateway filter chain
• GitHub Actions – for CI/CD automation and test pipelines

An ETL workflow framework on Macrometa GDN that automates data extraction, transformation, and loading between document collections and external sources. It supports log-based CDC and batch ingestion, schema inference, transformations using C8QL, and enables real-time data sync across Oracle, MySQL, PostgreSQL, MongoDB, Snowflake, and more.

Key responsibilities:

• Designed and implemented the workflow management and controller layer for Connection Workflows, including all external APIs for the controller that enable users to configure, trigger, and monitor data connector workflows.
• Contributed to the Meltano-based workflow execution and Argo Workflows orchestration layer.
• Contributed to or owned many of the available connectors.
• Contributed to GUI development and debugging as needed.

Languages: Java, Python
Tools:

• Macrometa GDN – storing data and executing transformations using stream workers
• Oracle, MySQL, MongoDB, PostgreSQL – syncing records with each database
• Meltano – ETL definition framework
• PyPI – hosting connector packages: https://pypi.org/search/?q=macrometa-&o=
• Argo Workflows – workflow orchestration engine
• React – implementing required GUI components for the GDN console

An AI agent designed to automate web GUI testing and reduce repetitive QA work on frontend applications. Built as a Crew AI multi-agent system, it interprets feature requirements and QA scenarios, then generates Cypress tests. The process starts by providing clear test documentation and a live app URL to browser-use, which identifies the necessary actions. Another agent then converts these steps into a Cypress test suite.

Key responsibilities:

• Led and developed this tool as part of a broader framework for automating QA, collaborating with other team members on the overall system.

Languages: Python
Tools:

• Browser-Use – External API contract design and validation
• CrewAI - Multi agent framework

A cloud-native, developer-centric API gateway for microservices that simplifies composing, securing, and monitoring microservice endpoints as managed APIs. It supports OpenAPI-driven configuration, runtime service discovery, legacy-to-modern API transformation, and can be deployed as lightweight, stateless containers in CI/CD pipelines.

Key responsibilities:

• One of the founding engineers
• Drove the architecture and initial implementation of the Microgateway core and toolkit.
• Defined the OpenAPI parsing engine, built the security and other API QoS features.
• Led containerization (Docker) and Kubernetes deployment strategies, established CI/CD workflows, and collaborated with UX and DevOps teams to expose a unified developer experience and rich observability.
• Took ownership of managing the open-source community around the project.

Languages: Java, JavaScript, Bash
Tools:

• OpenAPI Spec – for declarative API definitions and gateway config generation
• Docker – to package the gateway runtime as lightweight containers
• Kubernetes – for orchestrating and scaling gateway instances
• Prometheus, Jaeger, Zipkin – for collecting metrics and tracing
• JMeter – to load-test gateway performance under high throughput
• GitHub Actions – to automate build, test, and release pipelines

A Flutter-based cross-platform mobile application (iOS, Android) for browsing, viewing, and listening to Buddhist audio and video sermons. The app integrates with the existing backend supporting the website https://waharaka.com. Users can browse sermons by tags or keyword search. It includes a built-in audio player with seek, play, and pause functionality, as well as background playback. Users can also download audio sermons to their devices. Video listings are served via YouTube APIs, redirecting users to the YouTube app for playback. The app currently has over 5k downloads and a 4.9 rating from more than 270 reviews.

Key responsibilities:

• Responsible for approximately 90% of the application’s requirements gathering, design, and development.
• Currently working on version 2.0 in my free time, focusing on enhancing the user experience and adding new features.

Languages: Dart
Tools:

• Flutter – Cross-platform app development framework for Android and iOS
• YouTube APIs – To display live and uploaded video lists
• Android and iOS Developer Consoles – For publishing the applications

A fully featured API management solution including security, rate limiting, API analytics, and developer portals. The gateway component could handle 2500 TPS per instance while all other QoS features were turned on.

Key responsibilities:

• Contributed as a full-stack engineer throughout most aspects of the project (Specially around auth, rate limiting, subscription management, workflow management, publisher and developer portals).
• Visited multiple customers as an onsite product specialist.
• Mentored many juniors and interns.

Languages: Java, React
Tools:

• MySQL, PostgreSQL, MsSQL – as the data store for all API management components
• OpenAPI, WebSocket, GraphQL – support for all types of APIs and their governance
• JMeter – to load-test gateway performance under high throughput