Hello!
I'm Pubudu Jayakody

Gallery

Design and Development

The system is built with Unreal Engine as the visual and logic core, handling real-time rendering, materials, lighting, and scene interactions with high fidelity. All visual output is generated directly in Unreal, ensuring consistent quality across devices.

React+Vite is used for the frontend, providing a fast, responsive, and modular UI. User interactions are routed through Epic's Pixel Streaming plugin, which streams Unreal's output to the browser and returns user input back to the application in real time.

AWS hosts the entire deployment and uses a specifically engineered middleware layer to manage traffic flow, instance lifecycle, and on-demand scaling. This ensures stable performance, efficient resource usage, and the ability to support multiple users without requiring powerful local hardware.

● Visualization

► Assets: Assets were created as needed using Blender and Houdini, with additional elements sourced from Megascans and the Unreal Marketplace.

► Shader Development: Shader workflows were engineered with a strong focus on performance, leveraging compute shaders where beneficial to offload work from the streaming servers. Materials were kept highly procedural to support real-time parameter adjustments driven directly by the configuration system.

► Lighting & Rendering: Lighting and rendering were driven primarily by Lumen with hardware ray tracing enabled for maximum visual quality, while ensuring full compatibility and stable performance on Vulkan.

► Configurator Scripts: All scene interaction logic was developed to support real-time configuration, including systems for updating assets, materials, and layout changes instantly. This layer also handles the structured data exchanged between Unreal and the frontend.

► Binary Deployment: The final build was packaged for Linux focusing Vulkan, optimized specifically for cloud-based Pixel Streaming servers to ensure reliability and high performance at scale.

● UI/UX (Frontend) Development

The UI/UX layer was implemented as a standalone web interface built with React with Vite, providing full flexibility in layout, styling, and user interactions beyond what Unreal Engine’s native UI systems allow. This web interface communicates with the configurator through a WebSocket connection, exchanging JSON data via the Pixel Streaming plugin to trigger real-time updates inside the Unreal visualization layer.

By decoupling the UI from Unreal and delivering it through a web interface, the system enables greater design freedom, smoother iteration, and a cleaner, more modern user experience, while still maintaining fast, responsive, real-time control over all configuration features.

● Server Deployment

The system is fully deployed on AWS, supported by a purpose-built middleware layer that manages user flow and server lifecycles. As users join, the middleware allocates them to available compute instances, and when they disconnect or become inactive, those instances are transitioned into a warm pool rather than fully shut down, reducing operational costs while maintaining startup times under 30 seconds for new sessions. Throughout each session, key usage metrics such as activity duration and interaction patterns are captured to drive ongoing optimization.

The deployment is built on several core AWS components, EC2 for hosting the real-time Pixel Streaming servers, Lambda for queue logic and metrics processing, Redis for real-time tracking of user activity and inactivity, and DynamoDB for persisting user-related data and session records. Together, these services form a scalable, performant, and cost-optimized backend for deploying complex real-time applications.

Architecture Diagram - Simplified

Architecture Diagram - Detailed

● Used Tools

Unreal Engine Houdini Blender Quixel Megascans AWS React Python C++

Progress

Gallery

Design and Development

The system is powered by Unreal Engine, which serves as the core visual and logic framework. It manages real-time rendering, advanced materials, dynamic lighting, and scene interactions with high fidelity. All visual output is generated directly within Unreal, ensuring consistent, high-quality performance across devices.

The frontend interface is built with React+Vite, delivering a fast, responsive, and modular user experience. User interactions are transmitted through Epic Games' Pixel Streaming layer, which streams Unreal's rendered output to the browser while relaying user input back to the application in real time.

The entire solution is hosted on AWS, supported by a purpose-built middleware layer that orchestrates traffic management, instance provisioning, and on-demand scaling. This architecture ensures reliable performance, optimized resource utilization, and seamless support for multiple concurrent users, without requiring high-performance hardware on the client side.

● Visualization

► Assets: Assets were processed as needed using Houdini, with additional elements sourced from Megascans and the Unreal Marketplace.

► Shader Development: Shader workflows were developed with a strong emphasis on performance, leveraging compute shaders, particularly for the ocean simulation and its related components wherever advantageous to minimize load on streaming servers. Materials were designed to remain procedural, enabling real-time adjustments through the configuration system without requiring recompilation or heavy processing.

► Lighting & Rendering: Lighting and rendering were powered primarily by Lumen, with hardware ray tracing enabled to achieve the highest possible visual fidelity. The pipeline was carefully optimized to ensure full Vulkan compatibility while maintaining stable, consistent performance.

► Configurator Scripts: The interaction layer was designed to fully support real-time configuration, enabling instant updates to assets, materials, and layout changes without interrupting the experience. It also manages the structured data exchange between Unreal Engine and the frontend, ensuring seamless synchronization between the visual environment and the user interface.

► Binary Deployment: The final application was packaged for Linux with Vulkan, specifically tuned for Linux-based infrastructure to maximize reliability, scalability, and consistently high performance under heavy concurrent load.

● UI/UX (Frontend) Development

The UI/UX layer was developed as a standalone web application built with React, enabling complete flexibility in layout, styling, and interaction design beyond the constraints of Unreal Engine’s native UI framework. The interface communicates with the configurator through a WebSocket connection, exchanging structured JSON data via the Pixel Streaming plugin to trigger real-time updates within the Unreal visualization layer.

By decoupling the user interface from Unreal and delivering it as a web-based application, the system achieves greater design freedom, faster iteration cycles, and a more refined, modern user experience, while maintaining responsive, real-time control over all configuration features.

● Server Deployment

The system runs entirely on AWS, orchestrated by a custom middleware layer that governs user flow and server lifecycles. As users connect, the middleware assigns them to available compute instances, when they disconnect or go inactive, those instances are moved into a warm pool instead of being fully terminated, cutting operational costs while still enabling sub‑30‑second startup times for new sessions. During each session, metrics such as duration and interaction patterns are continuously recorded to inform ongoing tuning and optimization.

The deployment relies on a set of core AWS services, EC2 to host the real-time Pixel Streaming servers, Lambda to handle queuing logic and metric processing, Redis for real-time tracking of user activity and inactivity, and DynamoDB for durable storage of user data and session records. In combination, these components provide a scalable, high-performance, and cost-efficient foundation for complex real-time application deployments.

Architecture Diagram - Simplified

Architecture Diagram - Detailed

● Used Tools

Unreal Engine Houdini Quixel Megascans AWS React Python C++

Progress

Gallery

Design and Development

This project implemented a custom scattering system developed within Houdini to control plant distribution and growth behavior, and Houdini KineFX was utilized for animations, enabling procedural motion and natural deformation across various plant types.

● Assets Creation

► Several plant models (Orchid Flowers, Stems, Leaves and Roots) were fully procedurally generated, directly in Houdini using geometry nodes and attribute driven systems.

► Additional assets (Moss, Small Flowers and Plants, Tree Branch) were sourced from Quixel Megascans to enhance realism and variation.

● Look Development

Shading and rendering were performed using Redshift, for efficient look development and final outputs.

● Post Processing

Post-processing and compositing were handled in DaVinci Resolve for color grading and Nuke for precise visual adjustments and layering.

● Used Tools

Houdini Redshift DaVinci Resolve Nuke Quixel Megascans

Progress

Gallery

Design and Development

This project involved a full 3D production pipeline, including character modeling, terrain creation, vegetation generation, prop development, and FX simulations. Utilizing tools such as ZBrush, Blender, Marvelous Designer, Gaea, SpeedTree, and Houdini. Texturing was completed with Substance Painter and Designer. Look development and rendering were achieved with Redshift through a USD-based Solaris workflow for flexibility and efficiency. Final color grading, compositing, and visual refinements were carried out in DaVinci Resolve and Nuke. To streamline production and handle the large volume of assets, custom Python scripts were integrated into Houdini, significantly optimizing the overall workflow.

● Assets Creation

► Character Modeling: ZBrush and Blender were used for base modeling, Marvelous Designer for clothing, and Houdini's grooming tools for hair.

► Terrain Sculpting: Gaea handled the initial terrain sculpting and masking, while Houdini's terrain tools were used for fine-tuning and additional masking.

► Props: Small props primarily came from Megascans and KitBash, with some custom modeling done in Blender and Houdini as needed.

► Simulations & FX: Houdini was used for all simulations, including FLIP fluids, particle effects and etc.

► Vegetation: SpeedTree and Houdini were used for creating vegetation assets.

► Texturing: Substance Painter was used for asset texturing, while Substance Designer was employed to create custom tiling textures when required.

● Look Development

Redshift was used for all shading and rendering, while Houdini Solaris, leveraging a USD based workflow, enabled fast and flexible look development.

● Post Processing

DaVinci Resolve was used for color grading during post-processing, with Nuke handling precise compositing and layering to achieve the final visuals.

● Automation

To manage the large volume of 3D assets, textures, and other elements, custom Python scripts were integrated with Houdini, significantly accelerating the workflow.

● Used Tools

Houdini Blender ZBrush Marvelous Designer Gaea SpeedTree Substance Painter Substance Designer Redshift Python DaVinci Resolve Nuke Quixel Megascans KitBash3D

Progress