Sangeetha Suresh

Built machine learning pipelines to estimate above-ground tree biomass from high-resolution satellite imagery.

Processed and analyzed over 10 TB of geospatial raster data using python on high-performance computing (HPCC) resources.

Developed, trained, and evaluated deep learning models including CNNs, U-Net, Transformers, Attention-based architectures, DeepForest, and YOLO for remote sensing and geospatial analytics.

Collaborated with sustainability scientists and stakeholders to translate carbon science research into operational, map-based decision-support tools.

Designed and deployed interactive visualization applications to communicate carbon analytics and geospatial insights to technical and non-technical audiences.

Led a 4-day technical training seminar for international collaborators in Delhi, India (December 2024), providing instruction on geospatial data processing and analytics workflows.

Managed and expanded laboratory computing infrastructure, serving as technical lead for hardware, software, and HPC resource administration.

Performed advanced geospatial image processing and analysis using ERDAS IMAGINE and ArcGIS.

Simulated and analyzed the trajectories of chaotic mechanical systems eg. double pendulum.

Modeled and predicted system evolution using data-driven methods such as Dynamic Mode Decomposition (DMD) and Proper Orthogonal Decomposition (POD).

Implemented simulations and analyses using Python, with NumPy and Matplotlib for computation and visualization, and Git for version control.

• Developed an experimental framework to automate the evaluation of student responses using Large Language Models (LLMs).

• Evaluated and compared multiple LLMs on accuracy, consistency, efficiency, and reliability in educational assessment tasks.

• Designed and tested prompt-engineering strategies to reduce hallucinations and improve model performance across diverse response types.

• Built an automated evaluation pipeline to compare model-generated feedback with human-graded responses using quantitative and qualitative metrics.

• Analyzed model bias, consistency, and error patterns to improve robustness, fairness, and scalability of AI-assisted assessment systems.

• Presented research findings at the 2024 AERA Annual Meeting.

• Developed and simulated Shor’s and Grover’s algorithms using Qiskit on IBM’s quantum computing platforms.
• Gained hands-on experience with quantum gate design, low-level hardware simulation, and execution on cloud-based quantum systems.
• Implemented experiments using Python and Jupyter notebooks with Qiskit for quantum circuit development and analysis.