Atishay Narayanan
Math student at Princeton working on cooperative game theory and quantum machine learning. Devoted to pushing the frontier at the intersection of math and computer science.
Skills
Education
Princeton University
Sep 2023 – May 2027B.A. Mathematics, Minors in Computer Science & Classics · Princeton, NJ
Work Experience
Quantum Machine Learning Researcher
May 2025 – Jul 2025SenSIP Lab, Arizona State University · Tempe, AZ
- —Developed Quantum-Classical hybrid GPT models for image generation using PyTorch and Pennylane.
- —Demonstrated feasibility of quantum self-attention for transformer models.
- —Created one of the first applications to integrate a Quantum Mixed-State Attention Network.
- —Observed a 10% increase in PSNR compared to classical baseline models.
Teaching Assistant
Sep 2024 – PresentPrinceton University Computer Science Lab · Princeton, NJ
- —Teach 250+ new CS students algorithms and data structures concepts in Java.
- —Guide students through object-oriented design principles.
- —Assist students with debugging and project architecture.
Investment Associate Intern
May 2024 – Jul 2024SPHERE Investments · Miami, FL
- —Designed an exhaustive database of 1000+ potential limited partners for a healthcare real estate asset management firm.
- —Surveyed international clinical services and formulated an extensive report on investments into patient care hotels.
- —Conducted due diligence with healthcare executives to evaluate hospital resource allocation and patient outcomes.
Research
Weighted Shapley Value Without Symmetry or Exogenous Weights
Atishay Narayanan & Faruk Gul
Mathematics Junior Independent Work, Princeton University
An axiomatization of the weighted Shapley value that avoids assuming symmetry or exogenously supplied weights. Using Efficiency, Consistency, Null Player, and Linearity on totally monotone cooperative games, I prove that any conforming allocation is the weighted Shapley value for some endogenously determined weight vector — with the proof proceeding via the Möbius transform.
Quantum-Classical Hybrid GPT Models for Image Generation
Atishay Narayanan, Gennaro De Luca, and Andreas Spanias
SenSIP REU, Arizona State University
Developed Quantum-Classical hybrid GPT architectures for image generation, integrating a novel Quantum Mixed-State Attention Network. Demonstrated the feasibility of quantum self-attention within transformer models and achieved a 10% improvement in PSNR over purely classical baselines using PyTorch and Pennylane.
Advantages of Confirmation Bias in Bayesian Inference
Atishay Narayanan
PSY 360 Research Project, Princeton University
Designed and simulated Bayesian agents with asymmetric learning rates to model confirmation bias in learning stationary and non-stationary Bernoulli distributions. Evaluated agents on convergence speed and accuracy via MSE over 1000+ trials. Demonstrated that low-degree confirmation bias (10–25%) improves short-term convergence but degrades long-term accuracy, showing that its advantages are time-scale dependent.
Projects
A full-stack research paper tracking app with AI-powered relevance scoring via embeddings and semantic recommendations. Integrates with the Semantic Scholar and OpenAI APIs.