I will build and deploy computer vision and vision language models

Turn your visual data into production-ready AI applications. I specialize in traditional computer vision (detection, tracking, segmentation) as well as state-of-the-art Vision-Language Models (VLMs) for multimodal reasoning, OCR, and document understanding.

Whether you need real-time YOLO tracking or a fine-tuned VLM served as a low-latency API, I deliver optimized end-to-end solutions.

What I Offer:

• Vision-Language Models (VLMs): Fine-tuning and serving Qwen2.5-VL, Florence-2, and LLaVA for visual Q&A, document AI, and zero-shot reasoning
• Real-Time Detection & Tracking: Custom YOLO (v10/v11), Detectron2, and OpenCV pipelines
• Segmentation & OCR: Instance segmentation (SAM) and complex document text extraction
• VRAM & Speed Optimization: Model quantization (AWQ, GGUF), TensorRT conversion, and vLLM/vLLM-vision serving
• Production API & Deployment: FastAPI endpoints, Dockerization, and serverless GPU hosting (RunPod, Modal, AWS, Hugging Face)

Supported Frameworks & Models:

Qwen2.5-VL, Florence-2, YOLO, PyTorch, OpenCV, vLLM, Roboflow, FastAPI, Docker, CUDA/TensorRT.

Please message me before ordering to discuss your dataset, VRAM budgets, and target latency requirements!