Thiwanka Jayasiri

• Self-motivated, with a strong sense of personal responsibility.
• Excellent communication skills, both verbal and written.
• Proven ability to learn quickly and adapt to new situations.
• Skilled at working independently and collaboratively in a team environment.
• Worked well in a team setting, providing support and guidance.
• Hands on Systems Architect/Engineer delivers STOA AI models for warehouse Health & Safety tracking, On-Time-Performances, etc.
• Built a custom automated training pipeline for new customer onboarding
• Build Generative AI pipelines for reporting, and consolidating Health & Safety reports, optimizing ETL workflows.

• Researching on Safer AI infrastructure, and LLM Evals.
• Participated in peer review processes for academic journals, providing constructive feedback on manuscripts submitted by fellow researchers.
• Published numerous whitepapers in AI safety forums/journals, which have been widely cited by other scholars in various AI disciplines.
• Conducted comprehensive literature reviews to identify gaps in existing knowledge and inform future research directions.

• Oversaw developing and managing the data product pipeline, ensuring alignment with strategic goals and technological best practices.
• I mentored less experienced team members, fostering professional growth and enhancing team capabilities while managing relationships with key business stakeholders.
• I actively engage with the Data Engineering Chapter and GenAI Squad, leading product development and engineering efforts to bolster Air New Zealand's Customer Success Division.
• I played a pivotal role in developing and architecting chatbot and GenAI solutions across various chapters of Air New Zealand, driving the successful production of these solutions through pilot projects and comprehensive stakeholder engagement, including domain experts and digital leadership.
• I have a comprehensive technology stack, including Snowflake, AWS Java SDK, Azure DevOps, GCP, Dbt, and Airflow, to drive innovation and efficiency in data management and processing.
  Reduced production costs with innovative engineering solutions, resulting in improved profit margins.
• I contributed towards sustainable operations by suggesting eco-friendly modifications in existing systems.
• I developed comprehensive documentation for new products, supporting efficient knowledge transfer throughout the organisation.
• I established strong relationships with key industry experts, which led to valuable collaborations on cutting-edge projects.

• Oversaw developing and managing the data product pipeline, ensuring alignment with strategic goals and technological best practices.
• provided mentorship to less experienced team members, fostering professional growth and enhancing team capabilities while also managing relationships with key business stakeholders.
• Actively engaged with the Data Engineering Chapter, leading efforts in product development and engineering to bolster Air New Zealand's Customer Success Division.
• Played a pivotal role in developing and architecting chatbot and GenAI solutions across various chapters of Air New Zealand, driving the successful production of these solutions through pilot projects and comprehensive stakeholder engagement, including domain experts and digital leadership.
• Utilized a comprehensive technology stack, including Snowflake, AWS Java SDK, Azure DevOps, GCP, dbt, and Airflow, to drive innovation and efficiency in data management and processing.

• Ensured AI solutions were innovative but also robust and scalable, guiding projects from conceptualization to production.
• Developed custom models for object detection, tracking, depth estimation, and synthetic data generation, leveraging LLMs and vision transformers.
• Enhanced model training and deployment efficiencies by utilizing Nvidia's Metropolis and TAO tooling.
• Implemented AI model deployment on edge devices, primarily employing AMD Xilinx chipsets, to achieve efficient real-time processing with minimal latency.
• Created custom software modules to improve the performance of machine learning models during inference and post-processing stages.
• Managed to create and operate the synthetic data generation pipeline, enhancing data quality for training more resilient models.
• Orchestrated the transition of models from development to production phases, establishing CI/CD pipelines for adherence to scalability, maintainability, and performance benchmarks.
• Led cross-functional teams, promoting best practices in AI and computer vision and nurturing a continuous learning and innovation culture.
• Integrated AI systems with Databricks to augment insights and transparency, offering end-users enhanced visibility of data processed by AI systems.
• Collaborated with developers and engineers to develop technological solutions while keenly understanding business processes, needs, and objectives.

• Developed and maintained a computational pipeline for bio-fluids, climate modelling, and simulations, incorporating machine learning, open-source software, and deterministic accelerated hardware.
• Collaborated with GeorgiaTech and industry experts to accelerate the project's expansion, driving innovation and staying current with domain advancements.
• Conducted market research to identify applications for the pipeline and strategies to increase its impact and outreach, integrating Databricks and Tableau for a Ministry of Education project to support informed decision-making.
• Evaluated the technology's market fit, engaged with end-users and stakeholders for feedback, and refined the pipeline—managed projects, including integrating Azure Synapse and Power BI to create interactive dashboards, notably for SmartLife.
• Utilized a technology stack including Pytorch, Tensorflow, Numba, PyCuda, Databricks, Tableau, Azure ML, Synapse, and Power BI for optimal data processing and visualization.
• Partnered with SmartLife on an energy profiling pipeline for intelligent grid monitoring, including developing a SmartGrid monitoring clamp-based PoC.
• Implemented best practices for data management and quality assurance to ensure result accuracy and reliability.
• Documented and shared project findings through technical reports, presentations, and white papers on Surrogate Models for computational fluid dynamics, contributing to the scientific community.
• Developed a framework for accelerated numerical computation using SIMD instructions, achieving an 8X speed increase over conventional Intel Matrix Multiplication libraries.
• Led system event storming sessions to refine system design and development, focusing on customer requirements with reduced technical challenges.

• I led the development and strategic direction of a machine-learning proof-of-concept system that will enrich the entertainment experience for over 100 million users. This initiative involved streamlining the system architecture to efficiently link glove hardware design with data quality enhancements.
• Directed optimizing machine learning infrastructure to facilitate multi-hand tracking, significantly reducing model complexity by consolidating 45 models to 2.
• Spearheaded the creation of a proxy computer vision system, establishing a comprehensive ground truth data lake that bolstered the precision of machine learning models for detecting various hand movements. It utilized advanced technologies, including Intel RealSense, TOF cameras, and Oculus and Magic Leap devices, for effective prototyping and solution development.
• Fostered customer engagement through direct communication, gathering insights to inform priority areas for system improvements and enhancements.
• Collaborated closely with executive management to assess and select future project initiatives, aligning them with the organization's strategic milestones and scheduling demands.
• Unified team efforts towards achieving common objectives, demonstrating leadership in driving project advancement within a highly collaborative environment.
• Executed detailed analyses and calibrations of innovative glove technologies, delivering comprehensive reports and setting precise conditions for optimal functionality.
• Advanced data integrity by developing sophisticated Python-based data cleaning and profiling routines. Leveraged Docker technology to refine predictive modelling, resulting in improved inventory management and a 35% boost in production capabilities.
• Produced and disseminated refined data visualizations, effectively communicating the outcomes of analytical processes to stakeholders.
• Engaged in cross-functional collaboration with organizational stakeholders to ensure alignment with overarching software strategies and architectural frameworks, prioritizing effective problem resolution.
• Generated essential engineering and technical documentation to streamline collaboration and facilitate product development.
• Initiated and guided research programs that aligned with the organization's strategic objectives, contributing to the advancement of technological innovations.
• Provided support in filing provisional patents related to biomechanics and proxy validation systems, working closely with academic institutions such as the University of Auckland Biomedical Engineering Group.
• Demonstrated strong leadership throughout project lifecycles, from initial concept to successful completion, enhancing project management and team leadership skills.

• Conducted in-depth research on various crypto consensus algorithms, leading to developing a framework for a distributed machine learning pipeline tailored explicitly for the emerging needs of Web 3.0 gaming.
• Developed and maintained a real-time data pipeline for game event monitoring, emphasizing the system's reliability and scalability.
• Designed, developed, and maintained ETL pipelines using DataBrew, ensuring timely and accurate data processing for the Generative AI workstream.
• Constructed an end-to-end Proof of Concept/Minimum Viable Product for gamifying AI in the metaverse, leveraging Unity and cloud platforms for optimal performance.
• Managed the development process through the Scrum framework to ensure adherence to project timelines and milestones.
• Collaborated with various teams to identify, prioritize, and implement product features, enhancing user experience and driving business growth.
• Undertook research and analysis to identify trends and patterns in data, providing insights and recommendations for optimizing product performance and user engagement.
• Formulated planning criteria essential for integrating and enabling new technological advancements.
• Developed a federated data exchange framework using open-source tools and techniques.
• Maintained diligent technical documentation and implemented knowledge-sharing practices to ensure effective communication and knowledge transfer across teams.

• I developed an AI application for vision/radar-based fall detection, aiming to enhance patient monitoring in care houses with a focus on privacy and efficiency. My work involved deploying MM-Wave radar sensors on Raspberry Pi devices across select care houses in Switzerland and configuring them for non-invasive patient tracking with data anonymization for privacy.
• I trained a machine learning model to recognize fall patterns from radar data, employing TensorFlow Lite for real-time deployment on Raspberry Pi devices. For data management and real-time communication, I set up InfluxDB for storing radar tracking data and implemented MQTT protocol communication using Mosquitto as the broker.
• I integrated Fluentd logging on Raspberry Pis for device health monitoring and configured Grafana dashboards for real-time system monitoring. A key part of my role was ensuring secure data transmission with MQTT-based encryption.
• In collaboration with the Ministry of Education, I developed an energy profiling prototype for heat pumps using gradient boost techniques. This involved deploying machine learning models on Raspberry Pi-based hubs and utilizing MM-Wave radar data to track occupancy, and optimize the heat pump utilization.

• Implemented an Object detection/classification model with Pytorch Yolov5 and integrated it with the production model, along with a new implement algorithm to draw and extract the vector measurements from floor plans.
• Developed a software module specifically engineered to detect and measure verge and overhangs from an elevation floor plan, which solves and improved the floor plan estimation process by 50%.
• Developed tools to measure AI performances, e.g. FP, TP, mAP , etc with the annotated dataset. Compatible with different labeling formats.
• Successfully automated a data file analysis ( .gen to .csv) tool to compare AI and human-generated estimations. The format was complex to solve when I used different mathematical techniques to wrangle the data and produce a compatible file for processing.
• Resolved problems, improved operations and provided exceptional service to the end user.
• Created plans and communicated deadlines to ensure projects were completed on time.

• I specialized in the development, architecture, and prototyping of Computer Vision applications, targeting critical challenges in the mining industry
• I successfully integrated the hardware and software and built an optical Gas Imaging System,
• I built a site asset tracking and detection, an Embedded Face Detection Application , a Fire & Smoke Detection Module, and a Safety Detection System for worksite.
• In addition, I was instrumental in designing and deploying hardware platforms capable of running AI on the edge. This involved the use of various devices and components such as M-2 VPUs, Raspberry Pi Single Board Computers (SBCs), Jetson Nanos, and Coral TPUs. These platforms were critical in enabling efficient and real-time processing for the applications I developed.

• I spearheaded the automation of the Metasurface generative design, boosting workflow efficiency by 30% and enhancing design accuracy.
• My implementation of Generative Adversarial Networks for nanophotonics metamaterials led to a significant 25% improvement in material performance.
• Through rigorous validation and reformulation of engineering models, I ensured a 20% increase in model reliability, directly impacting product quality.
• My efforts in quantitative analysis and synthesizing nanophotonics research contributed to a 25% enhancement in research accuracy and a 20% faster knowledge transfer.

• Managed to develop computer vision product on edge. This was achieved within 4-8 weeks time frame, this included product design, software engineering and hardware engineering.
• Composed production-grade code to convert machine learning models into services and pipelines to be consumed at web and mobile scale.
• I designed and developed Vision and Speech Synthesized product: Tensorflow, Tensorflow Lite , Arm architecture, Multi-lingual capabilities. Prototype runs on a RaspberryPi based setup, with 9 hrs battery life.
• Worked with associated technology and software applications to manipulate and synthesize data for research projects.
• Utilized fundamental knowledge of state of art engineering applications, computational analysis tools, patents and industry trends to carry out advanced research.
• Tested troubleshooting methods, devised innovative solutions, and documented resolutions for inclusion in knowledge base for support team use.
• I conducted R&D based on the RISC-V Architecture, Secure Micro-Kernels -sel4, and low powered AI applications for surveillance applications.
• Conducted engineering and detailed experimental tests to collect design data and assist in research work.
• Performed validation and testing of engineering models to support adequacy and reformulated models.

Completed projects,

• Implemented a Channel Propagation machine learning model to optimally position a UAV in all terrain profiles.
• Middleware Development for Sensor Fusion. ROS & RTI-DDS. Co-Authored a Python library to handle certified Airband radio on a remote gateway.
• Designing Software Define Network (SDN)for airborne multi-layer communication links handling.
• Developed a visual obstacle avoidance systems for UAVs: NVIDA Jetson TX2 /J20 base board as a hardware platform, TensorFlow , TensorRT and object detection and avoiding model for small aircrafts.

Achievements and contributions.

• Successfully implement systems engineering procedure and introduce an open source system engineering software Capella to ease design process and implementation.
• Successfully completed Graduate Fellowship objectives in implementing predictive radio channel propagation analysis method for airborne mesh networks.
• Successful completion of Vehicle Systems Management middle-ware developer training.
• Successful completion of MANET Administrator Training conducted by Persistent Systems, USA.
• Improved product engineering life cycle 50%.
• Conducted regression testing, analyzed results and submitted observations to development team.
• Tested troubleshooting methods, devised innovative solutions, and documented resolutions for inclusion in knowledge base for support team use.
• Reviewed project specifications and designed technology solutions that met or exceeded performance expectations.

• I led a team of 12 experienced developers and six interns. We focused on creating a new and innovative retail mobile application that combined AI and Augmented Reality (AR) features for a specific retail sector. We worked with a robust technology stack that included Python, C++, Android, Vuforia, Unity, and Microsoft Azure.
• Using OpenCV and TensorFlow, we made significant advancements in face tracking and detection, improving user interactivity within the app by 25%. We also refined our project planning process, creating detailed work orders based on user stories, specifications, and product features. This helped us hit 95% of our initial project benchmarks, ensuring that we delivered a high-quality product on time.
• I conducted weekly data-driven planning meetings to maintain customer engagement and stay ahead of emerging market opportunities. By analyzing sales, customer feedback, and market trends, we increased user engagement by 20% and sales by 10% during the first quarter after launching the app.
• I also implemented a metrics-based strategy for continuous product evaluation, using advanced data modelling and analysis techniques to identify trends and forecast user behaviour. Our predictive models were refined by integrating machine learning systems and AI algorithms, enhancing data precision by 30%.
• My strategic and technological initiatives, particularly optimising face detection and classification capabilities, reduced operating costs by 15%. This laid the foundation for future growth and innovation in our product offerings.

• I was responsible for the overall project management of business IT projects. I led projects related to Maersk's business releases 5 to 7 and conducted pilots across India, Sri Lanka, and Bangladesh. I collaborated with the World Trade Organization, Customs Organization, and Sri Lankan authorities to implement intelligence platforms successfully.
• I have defined clear roles and responsibilities among team members at the outset of each new assignment, minimizing confusion related to task ownership.
• I streamlined project coordination processes by implementing effective communication and documentation strategies.
• I monitored ongoing project performance closely, making adjustments as needed to ensure the timely completion of milestones.
• I balanced multiple simultaneous project deadlines while maintaining the quality and timeliness of deliverables.