Software Engineering Machine Learning

Saurabh Padmakumar

Ho Chi Minh City, Vietnam +84-339**-**** ******************@*****.***

linkedin.com/in/saurabh-padmakumar/

Education

Royal Melbourne Institute of Technology Saigon, Vietnam B.Eng in Software Engineering (Hons) Expected May 2025 Relevant Coursework: Practical Data Science, Machine Learning, Big Data Engineering, Research Methods for Engineers, Engineering Design (Interdisciplinary Engineering) Certifications and Honors

2024 Microsoft ASEAN AI for Accessibility Hackathon, 1st Prize Recognized for groundbreaking AI solutions to enhance accessibility, competing among leading innovators. International Excellence Scholarship, RMIT University Awarded for academic and extracurricular excellence. Certificate of Experience, Phuong Hai Scientific Laboratory For significant contributions in AI research, software engineering, and agricultural technology solutions. Certificate of Recognition, Emerging Leaders Project 2024 Honored for mentoring and guiding students in leadership and academic excellence. Skills

Programming Languages: Python, Java, C++, C, Swift, Scala, SQL, NoSQL, HTML, CSS, JavaScript Libraries: Pandas, NumPy, Matplotlib, Scikit-learn, PyTorch, TensorFlow Frameworks and Developer Tools: React, Spring Boot, JUnit, Tailwind, Shadcn/ui, Linux, MATLAB, Git, Docker, Redis, Redux, VS Code, IntelliJ IDEA, Jira

Big Data Technologies: Hadoop, Spark, Kafka, Databricks Language Proficiency: English (Fluent), Hindi (Fluent), Malayalam (Native) Experience

Undergraduate AI-Engineer March 2024 – September 2024 Phuong Hai Laboratory Ho Chi Minh City, Vietnam

• Conducted in-depth research on advanced computer vision and deep learning techniques, reviewing state-of-the-art architectures such as YOLO, Detectron2, Fast R-CNN, and HuggingFace Vision Transformers for enhanced model performance.

• Fine-tuned and optimized pre-trained models using transfer learning and rigorous hyperparameter tuning, adapting them to project-specific datasets and achieving robust detection accuracy.

• Integrated optimized models into scalable, real-time inference pipelines using OpenCV, ensuring efficient processing and deployment in production environments.

• Developed automated data preprocessing and augmentation workflows such as histogram equalization and noise addition) to improve model generalization.

• Benchmarked model performance with quantitative metrics, achieving mAP50 scores of approximately 93.5% for YOLOv8 and 93.6% for YOLOv9, with precision and recall consistently above 90% on diverse validation datasets.

• Implemented containerized deployment solutions with Docker and utilized MLflow for experiment tracking, reproducibility, and streamlined model monitoring. Undergraduate AI-researcher November 2023 – June 2024 SightSense Ho Chi Minh City, Vietnam

• Led a multidisciplinary team in the development of SightSense, overseeing research, design, and implementation of innovative solutions in computer vision.

• Facilitated interviews with stakeholders to gather requirements and refine project goals based on user needs and feedback.

• Delivered impactful pitches to stakeholders and potential investors, showcasing technical achievements and practical applications of the solution.

Projects

Kafka Pipeline for Real-Time Stock, Weather, and Synthetic Data Integration

• Developed a scalable real-time data pipeline using Apache Kafka to integrate live stock market data (via Alpaca API), weather data (via OpenWeatherMap API), and synthetic data (via Faker API).

• Engineered a robust ETL pipeline for ingesting, processing, and storing data into Cassandra databases, enabling efficient analysis and reporting.

• Conducted correlation analysis and built data visualization dashboards using Python (Pandas, Matplotlib, Seaborn) to identify key relationships between market factors, global events, and stock prices.

• Automated the deployment and management of Kafka producers, consumers, and Cassandra nodes using Docker Compose, ensuring seamless integration and scalability.

• Demonstrated the impact of economic factors such as currency fluctuations and global events on Microsoft’s stock performance using scatter plots, heatmaps, and time-series visualizations.

• Key Takeaways: Highlighted the interplay between stock performance and macroeconomic indicators, providing actionable insights for investors and analysts. Gained expertise in real-time data streaming, distributed databases, and data visualization techniques, showcasing the ability to work with large-scale systems. Machine Learning Pipeline for Energy Consumption Forecasting

• Developed a robust PySpark pipeline to ingest heterogeneous energy consumption datasets (electricity and gas) from MongoDB, covering multiple companies (Stedin, Coteq, Westland-Infra). Engineered custom Spark Transformers to clean, standardize, and enhance raw data.

• Performed rigorous Exploratory Data Analysis (EDA) to identify schema discrepancies, analyze variance and skewness, and visualize data distributions—informing model selection and feature engineering strategies.

• Designed and implemented separate ML models for electricity and gas consumption using Spark MLlib’s Decision Tree and Random Forest regressors. Optimized models through hyperparameter tuning and evaluated performance using metrics such as MAE, RMSE, and R . Employed MLflow to log experiments, track hyperparameters, and manage model deployment for reproducibility and scalability.

• Generated detailed visualizations and performance analyses to compare model effectiveness across different companies, guiding strategic decisions on tailored model design. OpenEduCloud: Linux-Based Cloud Storage and Database Management for Educational Institutions

• Designed and implemented OpenEduCloud, a Linux-based prototype server to provide smaller educational institutions with a cost-effective alternative to proprietary cloud storage and database management solutions.

• Hosted a cloud storage solution using Nextcloud and a database management system via NocoDB, leveraging the stability and security of Linux for optimal performance.

• Deployed on a hardware setup featuring an M2 ARM-based chip, 28GB of storage, and 4098MB of RAM, with VMware virtualization for resource efficiency.

• Utilized the LAMP stack (Linux, Apache, MySQL, PHP) for server configuration and employed Docker for containerized applications, ensuring scalability and ease of maintenance.

• The project highlights the power of open-source Linux-based solutions in reducing costs while enhancing the learning and instructional capabilities of educational institutions. Flower Image Recommendation System Using Deep Learning

• Developed deep learning models to recommend 10 flower images similar to the input flower image provided by the user.

• Implemented three different approaches:

Model 1: K-Means Clustering – Employed unsupervised clustering to categorize images with similar visual patterns and content. After feature extraction via a base Convolutional Neural Network model, K-Means clustering identified clusters of akin images. By calculating cosine similarity between the input image and others in the same cluster, 10 similar images were recommended.

Model 2: Siamese Neural Network (SNN) – Utilized a deep learning architecture comprising two neural networks to acquire rich representations of images through comparison. Trained on image pairs with similarity labels, the SNN computed similarity scores between the input image and each dataset image. Following score sorting (excluding the input image), 10 similar images were suggested.

Model 3: CNN with Image Similarity Calculator – Integrated a CNN model for flower type classification with an image similarity calculator. Extracted color histograms, texture features, and contours from the input image, the CNN predicted the flower type. Leadership

Secretary-Treasurer, RMIT Student Council June 2024 – Present Finance Assistant (CFO in-training), RMIT NeoCultureTech April 2024 – June 2024 AI Mentor, RMIT University

Contact this candidate