Meet Monpara

M.S.

Computer

Science candidate specializing in computer vision and applied machine learning with 3+ years of hands-on experience developing neural network–based systems. Expertise in PyTorch, YOLO (v8–v11), CUDA, and TensorRT for end-to-end model development, optimization, and real-time GPU deployment. Experienced in dataset curation, cross-dataset validation, latency profiling (mean/P95/P99), and accuracy–throughput trade-off analysis for robotics and UAV perception workloads.

Junior Machine Learning Engineer - Toplo Technology

(2020-06 - 2023-08)

• Deployed a CNN/ResNet-based image classification system (PyTorch) with optimized inference performance, supporting efficient processing of ~50K+ images in production scenarios
• Implemented a GAN-driven synthetic data generation pipeline, reducing manual effort and saving ~$3.2K in testing costs while improving test coverage
• Designed scalable data pipelines (Pandas/NumPy) to support model training and inference, enhancing data reliability and pipeline efficiency

Machine Learning Graduate Researcher - Lamar University - Beaumont, TX

(2024-08 - 2026-05)

• Achieved 49.66% mAP@0.5 and 85.84% precision on the UAV-based autonomous perception dataset, outperforming YOLOv11 baselines.
• Proposed H-YOLO architecture with BiFPN-based multi-scale fusion and dual-attention detection head to improve tiny-object sensitivity.
• Performed cross-dataset validation on VisDrone for tiny-person detection, demonstrating robust generalization under domain shift.
• Built and annotated a real-world UAV flood dataset (508 images, 70/10/20 split) for tiny-person detection.
• Designed and benchmarked an end-to-end real-time inference pipeline, measuring latency (mean, P95/P99) and FPS on GPU systems
• Implemented evaluation and benchmarking utilities in C++ and Python for performance analysis
• Deployed and benchmarked models on NVIDIA GPUs using CUDA and TensorRT FP16
• Analyzed accuracy–latency trade-offs using GFLOPs, parameter counts, and FPS for real-time UAV deployment.
• Achieved stable real-time GPU throughput through end-to-end latency optimization and GPU-only post-processing.

Bachelor of Engineering - Electrical Engineering - Pandit Deendayal Energy University (2016-08 - 2020-11)

Master of Science - Computer Science - Lamar University (2024-08 - 2026-05)