Andrew Long

Designed and fine-tuned domain-specific generative AI models for tasks such as intelligent summarization, code generation, and automated customer support. Built RAG-based pipelines using LangChain, Weaviate, and Pinecone for enterprise search solutions; deployed scalable endpoints via FastAPI and serverless containers.

Designed and deployed OCR and object detection models (YOLOv5, Faster R-CNN) to automate medical device identification and traceability, reducing manual inspection time by 40% and improving labeling accuracy by 28%. Developed semantic segmentation models using self-supervised learning for microscopic images, enhancing the test IoU by 0.3 and implemented post-processing pipelines to refine predicti

Developed a custom BERT-based QA system for technical document retrieval and a root cause analysis model using anomaly detection, decreasing resolution time by 22% and boosting alarm handling accuracy. Engineered a pipeline to clean, embed, and index multi-lingual telecom documents, and created a feedback loop to improve QA model performance over time.

Led the development of a DL-based hybrid gaze tracking model to enhance VR user interaction, improving accuracy by 15% and enabling real-time integration in immersive applications. Collected gaze dataset using eye-tracking hardware, preprocessed noisy input data, and optimized inference speed for live feedback in Unity-based environments.

Built and deployed time series forecasting models (ARIMA, Prophet, LSTM) to predict food prices and purchase amounts, resulting in a 34% improvement in forecast precision and a 20% reduction in food waste. Built scalable ETL pipelines for real-time data aggregation from vendor APIs and inventory databases, enabling dynamic forecasts and inventory dashboards.

Pursued a Master of Science in Computer Science, focusing on advanced topics in the field. Gained in-depth knowledge and practical skills relevant to modern computing.