DILANKA SANULA

• Build and maintain B2B applications leveraging Large Language Models
• Lead the development of components and entire ML backend systems
• Developing stored procedures for data injection from lakes to warehouses and marts, specifically for populating chart data to support decision-making
• Generate synthetic data and fine-tune models to improve accuracy and efficiency of automated systems
• Implement advanced deployment pipelines for machine learning applications, utilizing modern CI/CD practices to enhance flexibility and scalability
• Deploy serverless architectures to efficiently manage scalable microservices, optimizing cost and performance
• Deploy backend systems using containerization, mastering the use of orchestration tools for improved scalability and reliability.
• Collaborated with cross-functional teams to integrate machine learning solutions into existing software systems, streamlining processes and boosting productivity.

Actively involved in a major project, ConverseAI, a conversational AI platform for Ecommerce.

• Preprocessing data for synthetic data generation and model training
• Developing serverless model training pipeline based on AWS Batch that reduced training time by approximately by 80%.
• Drawing insights from top research in academia and industry.
• Designing scalable architectures and data pipelines, and managing infrastructure development
• Deployment of conversational AI components using Amazon ECS, ensuring high availability and performance.

Primarily involved in the ‘Ryesight Coco' project, which automated the counting of approximately 300,000 coconuts on a conveyor belt. This project won recognition at the NBQSA National IT Awards Ceremony under the Industrial Supply Chain category. Detailed responsibilities included:

• Sensor-driven data collection, preprocessing, and augmentation to prepare for robust model training and fine-tuning to ensure high accuracy.
• Optimized system performance continually.
• Configured system on NVIDIA Jetson for high-speed influencing using TensorRT.
• Comprehensive Hardware Setup (Ryesight Coco project): Selected and configured cameras, edge devices, actuators, and lighting systems for optimal performance.