Nihar Sanda

Background

About Nihar Sanda

As a Machine Learning engineer with expertise spanning computer vision, NLP, and multimodal systems, I bridge the gap between cutting-edge research and real-world applications. My journey in AI has been defined by developing systems that genuinely impact people's lives—from emotion recognition systems that enhance educational experiences to privacy-preserving federated learning frameworks that protect sensitive data while maintaining exceptional model performance. My technical foundation includes deep expertise in PyTorch, TensorFlow, and transformers-based architectures (BERT, GPT, Vision Transformers), complemented by hands-on experience with production-ready ML infrastructure. At Vanderbilt University, I architected scalable real-time inference systems that significantly reduced latency while maintaining near-perfect uptime on Kubernetes, demonstrating my ability to deploy ML solutions that perform reliably at scale. I'm particularly passionate about multimodal AI systems that fuse computer vision and NLP techniques, as shown in my work developing affect recognition models that achieved remarkable accuracy improvements. My background in re-identification systems utilizing triplet loss and Kalman filters for trajectory estimation showcases my ability to solve complex tracking problems in uncontrolled environments. Beyond technical skills, my collaborative approach has been recognized through prestigious awards and open-source contributions to projects at organizations like CERN and the PEcAn Project. My published research in ACL Anthology and Taylor & Francis demonstrates my commitment to advancing the field of AI through rigorous academic exploration. I'm excited to connect with forward-thinking teams where I can leverage my expertise in computer vision, NLP, and large language models to build intelligent systems that solve meaningful problems at scale.