I’m Juan Martin Maidana, a Backend and ML Engineer at Silver River Software in Uruguay, where I architect enterprise-scale AI systems for industrial equipment management. Currently in my final semester at Universidad Católica del Uruguay, I’m completing my thesis on “Strategies for Integrating Computer Vision Models in Real-World Environments” - exploring optimal deployment paradigms for CV systems on edge devices like NVIDIA Jetson.

What I Do

At Silver River, I specialize in building production AI infrastructure for heavy machinery dealers and operators. My work involves designing multi-modal document processing pipelines that combine computer vision (YOLO), NLP (GPT-4, LangChain), and custom ML models to digitize technical manuals and parts catalogs - transforming 100K+ pages of legacy documentation into searchable, intelligent systems.

Beyond my professional work, I’m constantly experimenting with new technologies and approaches. I learn by building - whether it’s implementing YOLOv8 for wildlife tracking in Uruguay’s natural areas, creating real-time object detection systems, or exploring the latest computer vision architectures. Each project in my portfolio represents hands-on exploration of different ML paradigms and deployment strategies.

Technical Expertise

Languages: Python, JavaScript/TypeScript, SQL
ML/AI: PyTorch, TensorFlow, OpenAI APIs, LangChain, YOLO, Transformers, Vector Databases
Backend: Django, FastAPI, Celery, Redis, PostgreSQL, REST APIs
Infrastructure: Docker, AWS, Edge Computing (NVIDIA Jetson), MeiliSearch
Computer Vision: OpenCV, PIL, PDF processing, OCR, Image segmentation, Roboflow

Learning Through Building

I believe in learning by doing. My case studies showcase this approach:

• Built custom wildlife detection models using YOLOv8 with augmented datasets
• Implemented distributed processing systems with Celery for async AI workloads
• Created hybrid AI architectures combining Vision LLMs with statistical methods
• Deployed optimized models to edge devices with quantization and pruning techniques

My thesis research on edge computing deployment strategies and my industry work on cloud-based AI systems give me a unique perspective on the full spectrum of ML deployment options - from resource-constrained devices to distributed cloud architectures.

Current Focus

I’m passionate about bridging the gap between cutting-edge AI research and practical industrial applications. Whether it’s automating parts catalog digitization at work or tracking wildlife in Uruguay’s ecosystems as a personal project, I approach each challenge with the same philosophy: use the right tool for the job, prioritize robust engineering over complexity, and always validate with real-world data.