Experience

inspiring selection

AI Engineer — Trine University, Michigan

Smart Prompter (GPT-4 integration): Built an app that converts user keywords into optimized prompts; reduced manual prompt engineering effort by ~90%.
YouTube Summarization (Gemini LLM): Developed a Python/FastAPI pipeline integrating transcript APIs with Gemini to produce accurate, content‑ aware summaries.
Lyra AI Assistant (macOS): Created a lightweight personal AI assistant using a local LLM (e.g., via Ollama) and Python Tkinter; supports daily tasks and offline usage.

Built and deployed RESTful APIs for ML models using FastAPI and Docker, enabling integration into enterprise apps and dashboards at scale.
Applied NLP for chat log analysis, sentiment detection, and smart ticket classification using spaCy, Transformers, and BERT-based models.
Designed and implemented deep learning models (TensorFlow/PyTorch) for image/predictive tasks, achieving up to 94% accuracy.
Automated end-to-end training pipelines (feature engineering, hyperparameter tuning, model versioning) with MLflow and CI/CD, accelerating cycles by ~40%.

Built a CI/CD workflow for ML models using GitHub Actions, Docker, and MLflow for automatic retraining and deployment, reducing downtime by ~90%.
Deployed a real-time anomaly detection system (Autoencoders + Isolation Forest) for financial transactions, achieving 98% precision.
Implemented real-time pipelines with Python, Kafka, and TensorFlow Serving, reducing inference latency by ~45% and supporting high throughput.