Machine Learning Engineer | Deep Learning & MLOps
I am Pramit De, an M.Tech Computer Science student at Vellore Institute of Technology (VIT), specializing in High-Performance Deep Learning and Multimodal Representation Learning.
My journey into data science began with uncovering hidden patterns through exploratory data analysis—most notably identifying a hurricane event within 1.1 billion NYC taxi records. This experience shaped my approach to solving complex, real-world problems using data-driven and scalable AI systems.
Today, I focus on building edge-optimized AI systems and spatio-temporal models that deliver measurable impact in practical applications.
- Architectural Innovation: Improved multimodal alignment by integrating visual patterns with semantic and flow-based representations, achieving a 558% improvement in Macro-F1 (0.61 → 0.796).
- Edge Performance: Built an ONNX-optimized inference pipeline achieving 29.3 FPS on CPUs with a 28% reduction in latency.
- Structural Modeling: Designed a residual-driven GNN framework for irregular infrastructure networks, reducing forecasting error (MAE) by 54%.
- Prescriptive Outcomes: Developed a routing engine that reduced emergency response latency by 29.87%.
- Data Wrangling
- Statistical Modeling
- Machine Learning
- Deep Learning
- Analytical Thinking
- Business-Oriented Mindset
- Communication
- Adaptability
- Collaboration
- MLOps for scalable model deployment and monitoring
- Big Data Machine Learning for large-scale data processing
- Efficient and production-ready AI systems
- Email: pramitde726@gmail.com
- LinkedIn: de-pramit
- GitHub: Pramit726
- Portfolio: Portfolio
Always eager to collaborate on data science, machine learning, and real-world AI solutions.