I'm a Master of Information and Data Science candidate at UC Berkeley who loves to build AI systems! My work spans machine learning, software engineering, and data science – from creating scalable ML APIs to developing agentic systems that boost code generation accuracy.
- AI & Machine Learning: Building AI systems, natural language processing models, and edge-device optimizations.
- Software Engineering: Designing scalable systems with cloud platforms (AWS, Azure), container orchestration (Kubernetes, Docker), and robust APIs.
- Data Science & Analytics: Leveraging data and statistical theory to drive insights and optimize processes.
- Developing an AI-driven legal research platform, architecting an MVP using AWS Lambda, Bedrock, and Weaviate for strategic insights.
- Trained a lightweight CNN-BiLSTM model for non-invasive blood pressure estimation on edge devices with <5 mmHg MAE.
- Implemented terabyte-scale sensor data processing using TensorFlow, NumPy, and Pandas.
- Applied ML techniques with TensorFlow and Pandas on both structured and unstructured data to drive actionable insights.
- Built a sentiment analysis service using FastAPI and a pre-trained DistilBERT model.
- Deployed on Kubernetes (Minikube locally and Azure AKS in production) with automation scripts and Redis caching.
- Developed a multi-agent system using unit test-driven reflection to boost Python code generation performance across GPT-3.5, Claude, and Code Llama.
- Implemented a decoder-only Transformer model from Attention is All You Need, using NumPy for matrix operations and PyTorch for backpropagation.
- LinkedIn: linkedin.com/in/ankit-dey1
- Email: [email protected]