Suhrud Savargaonkar

Machine Learning Engineer @Everseen

London, GB

Fine-tuned, optimized, and deployed real-time computer vision models for theft detection at self-checkout, resulting in a 20% reduction in shrinkage and over 200% ROI across retail stores in Europe and the US.Developed a live-learning inferencing pipeline for real-time product-switch detection, deployed as a microservice to eliminate the need for full model retraining.Led an agentic AI system for intelligent data retrieval, enabling stakeholders to query real-time data from Elasticsearch using natural language.Designed and implemented LLM solutions for both log analysis and document summarization, resulting in automated test case verification and accelerated report generation, reducing manual effort from 2 days to 2 hours.Engineered and implemented a complete MLOps pipeline, encompassing data ingestion, model training, versioning, deployment,monitoring, and retraining specifically tailored for both Computer Vision and Large Language Models, ensuring seamless and reliable machine learning model lifecycle management in production.Technologies: Python, PyTorch, OpenCV, ONNX, TensorRT, NVIDIA Triton Inference Server, CLIP, FAISS, FastAPI, ELSER model, Hugging Face\'s Large Language Models (e.g, Pegasus), Docker, Podman, Azure Kubernetes Service (AKS), Azure Container Registry (ACR), Azure CI/CD, Elasticsearch, Convolution Neural Network (CNN), YOLO.