PraneetPraneetPraneet

Designed an adaptive maritime routing engine using CMEMS NetCDF datasets, modeling environmental interactions as dynamic cost fields on a spatio-temporal grid. Implemented a graph-based A* variant with efficient geospatial indexing (SciPy cKDTree) and large-scale data processing via xarray and NumPy for real-time, data-driven voyage planning.

Built a full-stack meeting-to-task automation system integrating LLM-based extraction with deterministic parsing and human-in-the-loop validation. Features include intelligent task breakdown, dependency detection, sprint management, and an inbuilt agent (“Nova”) for automated Scrum operations.

Developed an automated pipeline that generates, annotates, and trains image classification models entirely from text prompts using Gemini + Stable Diffusion and YOLOv8. Achieved high mAP with zero manual labeling, demonstrating scalable synthetic data workflows for computer vision model training.

Built a context-aware chatbot using pre-trained LLMs and a RAG framework to handle oceanographic and institutional data queries with factual precision. Integrated LangChain for pipeline orchestration, Chroma for vectorized retrieval, and Gradio for the user interface. The system efficiently retrieves and contextualizes domain-specific data, enabling enhanced understanding and conversational accuracy for scientific and administrative users.