Berkay Tor

Architected and implemented MappedGPT, an LLM-powered query interface using FastAPI and GraphQL API, integrating OpenAI GPT-4 to translate natural language prompts into precise GraphQL queries. Developed the retrieval pipeline using PGVector and Haystack, ingesting building metadata and query templates into PostgreSQL with vector embeddings for retrieval-augmented generation (RAG).

Led design and implementation of scalable NLP data pipelines (tokenization, lemmatization) on Databricks for the AdContext NLP Optimizer project, enabling downstream webpage classification, sentiment analysis, and entity recognition; leveraged PostgreSQL for structured text storage. Developed and fine-tuned transformer-based NLP models (BERT embeddings) on AWS (EC2/SageMaker) to enhance content re

Ensured HIPAA compliance in all development processes by implementing encrypted data pipelines and secure storage for protected health information (PHI), adhering to strict healthcare data privacy regulations. Developed computer vision models using TensorFlow CNNs to analyze medical imaging (X-rays, MRI), achieving ~90% accuracy in detecting diagnostic anomalies and accelerating clinical decision

Developed predictive ML models using Python (scikit-learn, TensorFlow/Keras) for image and signal data analysis, achieving over 90% accuracy on classification tasks and improving key performance metrics. Created RESTful APIs and web interfaces (Flask/Django back-end, React front-end) to serve ML models and present interactive data visualizations, enabling stakeholders to access AI-driven insights

Designed and implemented deep learning models for medical device diagnostics. Developed CNN-based image classification models (TensorFlow/Keras) to analyze clinical imagery, achieving high accuracy in detecting anomalies. Implemented real-time object detection using YOLO to identify issues in lab instrumentation, reducing manual review time by over 20%.

Developed image analysis and machine learning prototypes for diagnostic devices. Used ImageJ and MATLAB for medical image preprocessing and segmentation, applied principal component analysis (PCA) for feature extraction, and implemented prototype CNN models (TensorFlow/Keras) to classify sample data. Automated data processing tasks in Python to accelerate experimental analysis.

Completed a Master of Science in Computer Science, focusing on advanced topics and research within the field. Gained expertise in complex computer science principles and applications.

Obtained a Bachelor of Science in Computer Science, building a strong foundation in core computer science concepts. Developed fundamental skills in programming, algorithms, and data structures.