Gaurav Sawant

I’m an AI/ML Engineer with 1.5 years of experience in applied Machine Learning, Deep Learning, NLP, and multimodal systems—especially LLM fine-tuning and RAG pipelines. I publish research in Computer Vision and NLP, and I’m driven by building end-to-end AI solutions from research to deployment.

At The Unmarketing Agency, I built a multimodal Text Suggestion Engine using Vision-Language Models to generate titles, descriptions, and SEO tags from product images—reducing manual listing effort. I also implemented a hybrid search system (BM25 + semantic embeddings with FAISS) with tag-based filtering to improve search relevance.

On the computer vision side, I designed a two-stage pipeline combining YOLOv8 object detection with EfficientNetB0 classification, achieving 94% accuracy on 2000+ custom-annotated images for state condition identification. I additionally implemented an automated image quality validation pipeline so only compliant images flow downstream.

Earlier at Claysys Technologies, I developed a bank transaction fraud detection POC using a hybrid unsupervised/anomaly + sequential learning approach (Isolation Forest, Autoencoders, LSTM feature extraction) feeding Random Forest and XGBoost. In parallel, my projects like DocuMind and YTResponder strengthened my focus on evaluation, observability, and practical system design.