From ICLR spotlight papers to production-ready privacy infrastructure
Token-level differential privacy for LLM inference. Install via pip, integrate in minutes, achieve formal privacy guarantees.
Drop-in endpoint for sensitive data. Redaction, paraphrasing, and formal DP before hitting closed-box models.
Upload documents with sensitive data and see DP-Fusion in action. Real-time privacy pipeline demonstration.
Peer-reviewed publications at ICLR, NeurIPS, ACL, and more. Differential privacy, secure AI, multimodal learning.
Every feature in our tools is backed by rigorous academic research published at top-tier venues
Authors: Rushil Thareja, Preslav Nakov, Praneeth Vepakomma, Nils Lukas
A breakthrough framework enabling token-level differential privacy for LLM inference through parallel context fusion. Achieves formal (ε, δ)-DP guarantees while maintaining high utility across multiple modalities.
13+ papers spanning differential privacy, secure AI, education, NLP, and multimodal learning
Authors: Yulian Wu, Rushil Thareja, Praneeth Vepakomma, Francesco Orabona
Novel approach to RLHF with differential privacy guarantees for both offline and online learning scenarios.
Authors: Zhuohan Xie, Dhruv Sahnan, Rushil Thareja, et al., Preslav Nakov
Symbolic benchmark for evaluating chain-of-thought reasoning capabilities in financial domain applications.
Authors: Deep Dwivedi, Ritik Garg, Shiva Baghel, Rushil Thareja, Jainendra Shukla, Mukesh Mohania
Multimodal framework analyzing effectiveness of educational content through audio, visual, and textual features.
Authors: Rushil Thareja, Ritik Garg, Shiva Baghel, Deep Dwivedi, Mukesh Mohania, Ritvik Kulshrestha
Automated system for detecting prerequisite relationships between educational videos using NLP and multimodal analysis.
Authors: Rushil Thareja
Investigation of extreme sentiments in social media and their correlation with mental health outcomes.
Authors: Ritik Garg, Rushil Thareja, Manak Bisht, Manavjeet Singh, Sarthak Arora, Jainendra Shukla
Comprehensive database and benchmark for emotion recognition in Indian cinema with cultural context.
Authors: Rushil Thareja, Venktesh V, Mukesh Mohania
Automated extraction of concept prerequisites and dependencies from academic textbooks using NLP.
Authors: Rushil Thareja, Ritik Garg, Shiva Baghel, Deep Dwivedi, Mukesh Mohania
System for automated generation of educational videos from database content to enhance learning materials.
Authors: Ritik Vatsal, Shrivatsa Mishra, Rushil Thareja, Mrinmoy Chakrabarty, Ojaswa Sharma
Comparative study of physiological and psychological responses across VR and traditional gaming environments.
Authors: Raghav Mittal*, Rushil Thareja*, Mukesh Mohania
AI-powered travel recommendation system using personalization and contextual understanding.
Authors: Deep Dwivedi, Rushil Thareja, Ritik Garg, Shiva Baghel, Jainendra Shukla, Mukesh Mohania
Real-time multimodal analysis system for measuring student engagement in live classroom environments.
Authors: Rushil Thareja, Ritik Garg, Shiva Baghel, Deep Dwivedi, Mukesh Mohania
Automated video analysis engine predicting educational effectiveness through multimodal feature extraction.
Formal privacy guarantees for LLM inference, RLHF, and machine learning pipelines.
Privacy-preserving inference, federated learning, and secure model deployment.
Multimodal learning analytics, content generation, and engagement prediction.
Natural language processing, video understanding, and cross-modal learning.
Chain-of-thought reasoning, symbolic benchmarks, and verifiable AI systems.
Emotion recognition, sentiment analysis, and psychological response modeling.
Leverage our research in production. Open-source tools available now.