Passionate about responsible AI and building secure applications. Currently working on machine learning research and developing a platform for non-technical users to create safe, deployable apps.
I'm a passionate data science student with a strong focus on responsible AI and ethical technology development. My journey spans from building full-stack applications to conducting cutting-edge research in AI safety, machine learning, and model alignment. I believe in creating technology that is not only powerful but also safe and accessible.
My research focuses on responsible AI, machine learning safety, and ethical technology development. I am particularly interested in machine unlearning, model alignment, and developing AI systems that are transparent, fair, and beneficial to society.
Conducted comprehensive research on machine unlearning techniques, focusing on methods for selectively removing specific data influences from trained machine learning models while preserving overall model performance.
A framework for secure IoMT health monitoring, integrating encrypted data management with LLM-assisted explainability to ensure patient data privacy and clinician trust.
Currently conducting research on testing probing methodologies, red teaming techniques, and fine-tuning strategies for an Indian-based small language model specifically trained on JEE mathematics problems.
Participated in a competitive red teaming initiative focused on identifying vulnerabilities and potential security issues in the GPT-5 model.
A unique Kaggle competition project in the field of Artificial General Intelligence (AGI) that involved generating golf code (extremely concise code) to solve complex image-based puzzles.
Investigated how multi-modal signals (text, engagement, identity features) interact with internal platform features to shape automated content moderation outcomes.
Implemented a simplified version of NeuroReset using OLMo-1B and LoRA to explore forgetting specific information while retaining general knowledge. Evaluated using Membership Inference Attack (MIA) and Mini-MMLU benchmarks.
Democratizing app development by empowering non-technical users to build and deploy applications with confidence. Our platform ensures that every app built is secure by design, integrating automated security measures and responsible AI principles from the ground up.
Automated security checks baked into the development process.
Intuitive interface for non-technical founders and creators.
Leveraging advanced AI to assist in logic and design.