AI Engineer at Pubsilon

• Designed and developed an AI platform (Next.js, FastAPI, Docker, Redis) to democratize access to and interaction with AI models—no coding required. • Trained, fine-tuned, and evaluated an image classification model (EfficientNet) using metrics like Accuracy, Recall, Precision, F1-score, and Confusion Matrix, including activation and feature map analysis for ROI inspection. • Implemented an AI chatbot leveraging various LLM APIs (OpenAI, LLaMA, Claude, Gemini) and the RAG technique for document-based conversation. • Used Docker containers for deployment and Redis for optimized data access

Machine Learning Engineer (FYP) at INTELLCAP Corp.

• Reproduced a model from a research paper published in 2024 (the first human astrocyte detection research paper), implementing a customized Faster R-CNN with ResNet-50 + FPN backbone in PyTorch. Integrated a custom anchor generator with variable sizes and aspect ratios, and tuned RPN and ROI heads to better fit astrocyte morphology. Adapted and aligned COCOstyle bounding box annotations to immunohistochemically stained slides (GFAP, ALDH1L1). • Fine-tuned the model on the custom astrocyte dataset using standard data augmentation and hyperparameter tuning, and evaluated performance using COCO metrics (AP, mAP). • Also developed and fine-tuned ResNet-50 and EfficientNetB0 classification models in TensorFlow for skin cancer detection, using a labeled medical image dataset. • Reorganized and optimized a Flask-based web application to efficiently serve multiple machine learning models, improving code structure, scalability, and inference response times

Data Scientist at TEC FORGE

• Developed a data visualization dashboard using Power BI and Streamlit to analyze 85,000+ lines and 165+ columns of street lighting data, enhancing decision-making for smart city infrastructure. • Helped the team identify the perfect solution for their specific use case, tailoring tools and approaches to their needs. • Linked Power BI with data extracted from the database in Excel format and used Streamlit to visualize real-time data from an SQL database, enhancing decision-making capabilities. • Executed data manipulation and processing efficiently using Python and Pandas for ETL processes, along with data visualization with Plotly. • Authored a detailed report on deploying the Streamlit app on a server, ensuring scalability and usability for long-term operations. • Led a training session on data analytics using Power BI and Streamlit, helping the team optimize street lighting audits by deepening data insights for cities.