Artificial Intelligence/Machine Learning R&D Intern Feb 2022 - Jul 2023

• Utilized Faster R-CNN (Detectron2) models to classify spectrograms of 5G, LTE, WLAN, FM, GSM, and Bluetooth signals, achieving 98% categorical accuracy with data augmentations such as phase offset, reducing the signal to noise ratio and mixing signals. Experimenting with a novel patching technique using ConvNext/ResNet models for signal detection, deploying in real-time via ONNX with Keysights Spectrum Analyzer - FieldFox. Created a custom tool to annotate signal spectrograms, aiding the Spectrum Analyzer team in building a golden test set for ground truth.

Research Assistant Feb 2022 - Jul 2023

My role involves developing and designing software tools to support e-recycling techniques for iPads, iPhones and Apple Watches

• Researching e-waste recycling techniques, focusing on tracking and removing fasteners from electronic devices using computer vision strategies under the supervision of Prof. Matthew J Travers and Howie Choset

Senior Software Associate Engineer Feb 2021 - Jun 2021

• Delivered the integration between multiple micro-services to facilitate communications to complete an APN order successfully for the Wireless VPN Orchestration Domain Manager (wVPN ODM). Won the Dream Team Award for the Annual Quarterly Awards - Q4, 2021-2022 for successfully automating the orchestration of the first 3 APN orders in production and reducing the SLA by 5-7 days.
• Developed the Employee Resource Management Portal to assist resource planning, management, and allocation of resources of a company. It aimed to overcome the challenges of conflicting resource priorities and help accurately predict future requirements, thus improving profitability levels and streamlining costs.
• Engineered the PAC0410 Database Automation System as part of Wireless VPN mission’s development team. Responsible for developing API Capability for data consumption and transformed a 20-year-old Enterprise spreadsheet containing information on network components into a strategic DB by incorporating data consistency and security checks. Developed comprehensive MySQL scripts to replace legacy spreadsheet macros for network load assurance purposes.
• Collaborated on building the My Telstra Portal to view and manage mobile and broadband plans for customers, make payments, track data usage, and shop for tech devices and accessories.

Research Intern Jun 2020 - Aug 2020

• Researched Convolution Neural Networks for face recognition systems to understand their vulnerabilities, Histogram of Gradients (HOG) for detecting facial landmarks. Studied SLAM technologies and Kalman filters to implement on an autonomous donkey car vehicle. Worked on Calibration of Cameras, image processing using filters and edge detectors to extract useful data for image/video analysis

Student Intern May 2020 - Aug 2020

• Worked at LRDE-Laboratory for Electronics and Radar Development at Defense Research and Development Organization, India. Completed the project on the design of TWT based Transmission using Flyback converters based on given specifications for radar applications.

Marketing Intern June 2018 - September 2018

• Worked on on-ground push sales, digital marketing, and external collaborations to improve the company's reach of its services. Received a Letter of endorsement for exceptional performance in Marketing and Sales domains.

AWS- Certified Cloud Practitioner April 2023

The AWS Certified Cloud Practitioner validates foundational, high-level understanding of AWS Cloud, services, and terminology. This is a good starting point on the AWS Certification journey for individuals with no prior IT or cloud experience switching to a cloud career or for line-of-business employees looking for foundational cloud literacy.

Databricks Data Analyst October 2022

• The Databricks Certified Data Analyst Associate certification exam assesses an individual's ability to use the Databricks SQL service to complete introductory data analysis tasks. This includes an understanding of the Databricks SQL service and its capabilities, an ability to manage data with Databricks tools following best practices, using SQL to complete data tasks in the Lakehouse, creating production-grade data visualizations and dashboards, and developing analytics applications to solve common data analytics problems.

Telstra Security Champion Fundamentals May 2022

• Advocate for the adoption of top-tier security practices to ensure ongoing protection in both infrastructure and software delivery processes.
• Promote the implementation of secure coding standards among interdisciplinary teams.
• Support DevOps teams in merging their build processes with corporate-level Static Application Security Testing (SAST) tools.
• Collaborate with cybersecurity teams to enforce security compliance and conduct vulnerability and penetration testing on infrastructure and applications.
• Acquire knowledge about the Open Web Application Security Project (OWASP) and necessary code scans, including Coverity, Veracode, SonarQube, and HawkEye, to address vulnerabilities.

June 2022

• The DELF (Diplôme d'Etudes en Langue Française) exam is designed for individuals who wish or need to receive official certification of their French second language proficiency. They are official diplomas issued by the French National Education Ministry and are valid for life. it is required to acquire French nationality or to immigrate in order to study or work in a francophone country. Completed the A2 certification

Marketing Intern Feb 2023

• This microcredential provides a comprehensive overview of the foundations of data engineering - a set of infrastructure platforms and capabilities that allows organisations to acquire, transform, store and curate data, with the ultimate goal of realising value from data.
• About this microcredential, Taking a broad perspective, this microcredential helps participants from diverse backgrounds to be better informed when working with data engineering teams, or planning for data engineering as part of their team's projects or operations. Beginning with an overview of a typical data value chain, the microcredential then introduces data infrastructure and data pipelines, alongside examples of implementation technologies. A range of issues around data quality, security, monitoring and governance are explored, with the ultimate goal of demonstrating how data engineering helps organisations extract and realise value from their data assets.

May - August 2023

• Utilized Faster R-CNN (Detectron2) models to classify spectrograms of 5G, LTE, WLAN, FM, GSM, and Bluetooth signals, achieving 98% categorical accuracy with data augmentations such as phase offset, reducing the signal to noise ratio and mixing signals.
• Experimenting with a novel patching technique using ConvNext/ResNet models for signal detection, deploying in real-time via ONNX with Keysight’s Spectrum Analyzer - FieldFox.
• Created a custom tool to annotate signal spectrograms, aiding the Spectrum Analyzer team in building a golden test set for ground truth.

December 2023 - May 2024

• Engaged in developing Product Identification solutions. This project involves categorizing product images for Entropy by utilizing Homography techniques and employing the Grounding Dino algorithms to enhance the integration of images from multiple cameras.
• Worked in Cylab, Biometrics center under the supervision of Prof. Marios Savvides. This project is in collaboration with a company called Entropy.

December 2023 - May 2024

• Engaged in pioneering research on e-waste recycling methodologies, specializing in the identification and extraction of fasteners from electronic devices, guided by the expertise of Professors Matthew J. Travers and Howie Choset. This project is in collaboration with Apple.
• Trained Models such as YOLOv8, Faster-RCNN and InceptionNet to detect electronic components such as screws and rivets.
• Implemented Kalman Filters and DeepSort algorithms to precisely track the location of various components within the recycling pipeline, leveraging a multi-camera, multi-object tracking system

December 2023 - May 2024

• Developed a decoder-only GPT model for NLP tasks (summarization, question answering, sentiment analysis, and named entity recognition). Pretrained on 54GB OpenWebText, used tiktoken encodings, and implemented multi-head self-attention, feed-forward networks, and positional encodings in PyTorch. Achieved high accuracy in sentiment analysis and reasonable performance in others.

December 2023 - May 2024

• Implemented the redesign of AlexNet for enhanced performance on multicore CPUs and GPUs. Utilized OpenMP for CPU parallelism to achieve 250x speedup and CUDA for GPU acceleration to achieve 420x speedup by resolving bank conflicts.

December 2023 - May 2024

• Engineered a face recognition system by fine-tuning a ConvNext v2 convolutional neural network, coupled with a contrastive loss.
• Achieved a classification accuracy of 92% across 7000 classes, placing within the top 2% for face verification performance
• Implemented CNN’s including ResNet, Efficient Net and ConvNext from scratch, optimizing and evaluating them on the VGGface2 dataset
• Performed ablation studies to study the impact of various data augmentation techniques such as CutMix, Mixup and regularisation techniques such as stochastic depth, drop blocks, and label smoothing.

December 2023 - May 2024

• The goal of this project was to build and train an efficient neural network model in PyTorch to perform frame-level phoneme
classification on speech data. The model takes in mel-spectrogram features extracted from speech audio recordings and predicts the phoneme label for each frame.
• Achieved the highest test accuracy of over 88% in the associated Kaggle competition.
• Performed optimal architecture exploration and selection of model hyperparameters like layers, activations, optimizer through iterative experiments.
• Performed comparative analyses between ReLU, GELU, Leaky ReLU amongst non-linear activation functions and SGD, Adam, AdamW amongst optimizers.
• Hyperparameter Tuning: Systematically tune hyperparameters like learning rate, batch size, context size, regularization modes such as dropout and weight decay to optimize model performance.

October - December 2023

• Participated in developing a Teacher-Student Knowledge distillation technique to prune large, multilingual ASR models into efficient, compact monolingual versions for on-device use in low-resource languages.
• Implemented annotation generation for unlabelled data in Voxpopuli and trained attention-based models for languages like Maltese and Greek
• The increased training data for the student model increased performance over the Commonvoice Only model. The training data is still not sufficient to outperform the pre- trained form with no pretraining on multilingual data. However, fine- tuning both Whisper models results in improved performance. This shows that fine-tuning supervised data from SeamlessM4T was ef- fective for fitting Whisper-small to both Maltese and Greek.

October - December 2023

• Implemented signal processing algorithms to decompose High-Density Electromyography (HDEMG) data into constituent motor unit action potentials (MUAPs), utilizing bandpass and notch filtering techniques to enhance signal-to-noise ratio.
• Applied blind source separation techniques to identify unique motor units, accounting for physiological constraints such as interspike intervals and propagation patterns, enhancing the understanding of neuromuscular activations
• Applied blind source separation techniques to identify unique motor units, accounting for physiological constraints such as interspike intervals and propagation patterns, enhancing the understanding of neuromuscular activations

October 2023 - November 2023

• Preprocessing and Eigenface Computation Utilization of a dataset with 1071 grayscale face images (64x64 resolution). Implementation of image preprocessing in both MATLAB and Python, including conversion to double precision and unraveling images into vectors. Introduction to eigenfaces concept: faces approximated by linear combinations of eigenfaces.
• Computing Eigenfaces with Karhunen-Loève Expansion (KLE) Extraction of eigenfaces using Singular Value Decomposition (SVD). Transformation of eigenface vectors into image format for visualization. Analysis of the first eigenface and computation of reconstruction error as a function of the number of eigenfaces (up to 100).
• Computing ICA Faces with FOBI (FastICA-Based Blind Source Separation) Implementation of Independent Component Analysis (ICA) using FOBI. Comparison of ICA faces and eigenfaces in terms of basis vector orthogonality and reconstruction error characteristics. Exploration of the impact of different numbers of ICA components on the quality of the resultant ICA faces.
• Adaboost Implementation for Classification Development of an Adaboost training algorithm to combine weak classifiers into a strong classifier. Design of an Adaboost prediction function to classify new data samples.
• Training Adaboost with Eigenfaces and ICA Faces Rescaling images to 19x19 and recomputing eigenfaces and ICA faces. Training Adaboost models using eigenfaces and ICA faces as features, with varying numbers of weak classifiers and eigenfaces. Analysis of classification errors and comparison between eigenface and ICA face representations.
• Face Detector Application Application of the trained Adaboost model for face detection in a specific image. Implementation of non-maximum suppression to refine detection results. Evaluation of the model's performance based on Intersection-of-Union (IoU) values and total number of detected bounding boxes.

October 2022 - July 2023

• Delivered the integration between multiple micro-services to facilitate communications to complete an APN order successfully for the Wireless VPN Orchestration Domain Manager (wVPN ODM). Won the Dream Team Award for the Annual Quarterly Awards - Q4, 2021-2022 for successfully automating the orchestration of the first 3 APN orders in production and reducing the SLA by 5-7 days.
• Migrated the orchestrator backend from SeaStreet to Camunda which offers several advantages for scaling and improving effectiveness. Supported and activated new production orders.
• Supporting the integration of APIs with CSM, Magpie and IP Mobile Activation Teams

July 2021 - September 2020

• Engineered the PAC0410 Database Automation System as part of Wireless VPN mission's development team. Responsible for developing API Capability for data consumption and transformed a 20-year-old Enterprise spreadsheet containing information on network components into a strategic DB by incorporating data consistency and security checks. Developed comprehensive MySQL scripts to replace legacy spreadsheet macros for network load assurance purposes.

Jun 2020 - Aug 2020

• Collaborated on building the My Telstra Portal to view and manage mobile and broadband plans for customers, make payments, track data usage, and shop for tech devices and accessories.

Feb 2021 - June 2021

• Developed the Employee Resource Management Portal to assist resource planning, management, and allocation of resources of a company. It aimed to overcome the challenges of conflicting resource priorities and help accurately predict future requirements, thus improving profitability levels and streamlining costs.

Feb 2021 - August 2021

• Engineered a mobile sanitation robot to assist front-line workers and the general public in curbing the spread of COVID-19, leveraging MQTT for remote communications, Robot Operating System(ROS) for localization and mapping of the bot. Showcased research findings at the 2021 International Symposium of the Asian Control Association on Intelligent Robotics and Industrial Automation(IRIA)

Jun 2020 - Aug 2020

• Developed mechanical models of the smart cane and headgear subsystems on AutoDesk Fusion 360 and simulated the results on Proteus 8 software. Worked with Enable India, an NGO for the blind in India. Responsible for implementing the YOLO algorithm on the Coco-Dataset for object detection.

Feb 2020 - Aug 2020

• Implemented SVM algorithms on the generated wrist movement dataset from an IMU sensor to distinguish between tonic/clonic seizures as part of the Smart India Hackathon 2020.

Jun 2020 - Aug 2020

• Collaborated on building an interface that recognizes the user using template matching and allows them to control and navigate through the GUI using hand gestures using OpenCV and PyAutoGUI libraries. Made use of concepts such as ROI, thresholding, and background subtraction.

Jun 2020 - Aug 2020

• Researched Convolution Neural Networks for face recognition systems to understand their vulnerabilities, Histogram of Gradients (HOG) for detecting facial landmarks.
• Studied SLAM technologies and Kalman filters to implement on an autonomous donkey car vehicle.
• Worked on Calibration of Cameras, image processing using filters and edge detectors to extract useful data for image/video analysis.