Yash Patel

Home | Selected Papers | Mentoring | Projects | Blog

Hi! I’m an incoming research engineer at Harmonic with significant coding experience in C++, Python/PyTorch, OpenGL/GLSL, OpenCL, and Unity and research experience in uncertainty quantification, robust/convex optimization, rare-event detection, control theory, and PDE surrogate modeling. I am most excited about using ML for accelerating scientific discovery.

Harmonic Jan 2026 – Present

Research Engineer

Anthropic Nov 2025 – Present

Research Engineer, AI Safety

Working on alignment and reliability of frontier models.

University of Michigan Sep 2021 – Dec 2025

PhD in Statistics · Ambuj Tewari

My research focuses during my PhD were on principled uncertainty quantification, robust decision-making, and AI for Science. Given the importance of uncertainty in evaluating scientific hypotheses, my initial work primarily centered around one core question: How can we design principled uncertainty estimates for black-box models and use such uncertainty optimally for decision-making?

Waymo Jun 2025 – Sep 2025

Data Science PhD Intern, Simulations · Aman Sinha

Implemented an ADMM-based distributed convex optimization algorithm in C++ for importance sampling of rare events to achieve a 20x speedup in the simulations pipeline.

Bose Jan 2025 – Jun 2025

Machine Learning Research Co-op · Russell Izadi, Shuo Zhang

Implemented SAC and PPO methods for adaptive-FIR noise cancellation (PyTorch). Developed novel transformer-based approach for Wiener filter adaptation that outperforms FxLMS (10% dB reduction). Performed linear system identification and analyzed transfer functions to assess ML filtering.

Meta Jul 2018 – Sep 2021

Senior Software Engineer (IC5) · Albert Parra Pozo

At Facebook, I worked on a number of projects, generally in 3D rendering and reconstruction. Some highlights:

Designed and implemented novel real-time (72 FPS) novel dynamic object reconstruction algorithm for 300k+ vertex meshes in Unity HLSL/C# based on linear-blend skinning (LBS)
Implemented real-time (72 FPS) point cloud, dense mesh, and TSDFs (KinectFusion) scene reconstruction & rendering on HMDs & lenticular displays with C++/OpenGL/GLES/OpenCL.
Implemented deep learning model (PyTorch) and optimized via Qualcomm SNPE & QAT to run at 30 FPS on Qualcomm SoC for Portal platforms. Added translation support for quantized nodes in PyTorch-JIT to Caffe2.
Added distributed rendering with Docker, RabbitMQ, and Kubernetes to Manifold camera (code). Reduced depth estimation time by 30%.

Princeton University Sep 2014 – May 2018

A.B. in Mathematics · Matt Weinberg
Certificates in Applications of Computing, Statistics & ML

My interests over undergrad meandered through many areas. Some highlights:

HyperLoop Pod Design [Project Report]

Princeton University Project Founder & Lead, 2015-2017

2x Top 30 Team, International SpaceX HyperLoop Pod Design Competition

Deanonymizing Bitcoin Transactions: An Investigative Study On Large-scale Graph Clustering [Project Report]

Princeton University Senior Thesis, 2018

Tesla Autopilot Analysis [Project Report]
Neural Branch Predictor [Website] [Code]

Polymarket Jan 2018 – Jun 2018

Early-Stage Developer · Shayne Coplan

Worked on core pre-ICO development, integrating Bancor protocol liquidity and exchanges with the primary TokenDAO in Solidity (Truffle.js, testrpc, geth).

Amazon Jun 2017 – Aug 2017

Software Engineering Intern

Built Java Spring MVC debugging service for Kiva Picking Optimization team. Deployed globally via AWS (EC2, S3, SNS).

OpenLoop Jan 2015 – May 2016

Co-founder & Princeton University Lead

Co-founded a coalition of six top universities (OpenLoop), raised >$150,000, and built an 18 ft functional pod selected as one of 30 teams in the International SpaceX HyperLoop Pod Design Competition (Pod).

Columbia University May 2015 – Aug 2015

Research Intern · Abdulrahmen El-Sayed

Developed and simulated agent-based models of self-efficacy dynamics for sexual minority populations enrolled in exercise coach programs (code).

Princeton Plasma Physics Lab Jun 2013 – Mar 2014

Research Intern · Ilya Dodin

Developed FDTD numerical simulations in C++/Python of the Vlasov equation (reference) to study plasma evolution (video).

Selected Papers

My work has largely focused on developing methods with end-to-end statistical guarantees to create reliable machine learning systems and layering robust decision-making on top of such uncertainty estimates, especially for scientific applications. My projects largely split into three headings: uncertainty quantification methodology, robust decision-making, and AI for Science.

Uncertainty Quantification Methodology

Conformal Prediction for Ensembles: Improving Efficiency via Score-Based Aggregation [Code]
Neural Information Processing Systems (NeurIPS), 2025
Rivera, E.O.* , Patel, Y.* (* equal contribution), Tewari, A.

Variational Inference with Coverage Guarantees in Simulation-Based Inference [Code]
International Conference on Machine Learning (ICML), 2024
Patel, Y., McNamara, D., Loper, J., Regier, J., Tewari, A.

Robust Decision-Making

Conformal Contextual Robust Optimization [Code]
International Conference on Artificial Intelligence and Statistics (AISTATS), 2024 (Oral)
Patel, Y., Rayan, S., Tewari, A.

Conformal Robust Control of Linear Systems [Code]
In Submission
Patel, Y., Rayan, S., Tewari, A.

Non-Parametric Conformal Distributionally Robust Optimization
ICML Workshop on Structured Probabilistic Inference & Generative Modeling, 2024
Patel, Y., Cao, G., Tewari, A.

AI for Science

Continuum Transformers Perform In-Context Learning by Operator Gradient Descent [Code]
In Submission
ICLR AI for Accelerated Materials Design Workshop, 2025
Patel, Y.*, Mishra, A.* (* equal contribution), Tewari, A.

Operator Learning for Schrödinger Equation: Unitarity, Error Bounds, and Time Generalization [Code]
In Submission
Patel, Y.*, Subedi, U.* (* equal contribution), Tewari, A.

A Greedy PDE Router for Blending Neural Operators and Classical Methods [Code]
In Submission
Rayan, S., Patel, Y., Tewari, A.

Diffusion Models for Probabilistic Deconvolution of Galaxy Images [Code]
ICML Machine Learning for Astrophysics Workshop, 2023
Li, Y., Xue, Z., Patel, Y., Regier, J.

RL Boltzmann Generators for Conformer Generation in Data-Sparse Environments [Code]
NeurIPS Machine Learning in Structural Biology (MLSB) Workshop, 2022
Patel, Y., Tewari, A.

Scalable Bayesian Inference for Finding Strong Gravitational Lenses [Code]
NeurIPS Machine Learning and the Physical Sciences Workshop, 2022
Patel, Y., Regier, J.

Patents

Holographic Calling for Artificial Reality
US Patent App. 17/360,693
AP Pozo, J Virskus, G Venkatesh, K Li, SC Chen, A Kumar, R Ranjan, BK Cabral, SA Johnson, W Ye, MA Snower, Y Patel.

Mentoring

During my PhD, I have also had the opportunity to mentor the following fantastic undergraduate and master’s students on their theses and research projects.

Guyang (Kevin) Cao (Next step: Ph.D. in Computer Science at University of Wisconsin-Madison)
Honors Thesis, 2023-24
Undergraduate Research Program in Statistics, 2023
Non-parametric Conformal Distributionally Robust Optimization
Zhiwei Xue (Next step: Ph.D. in Computer Science at National University of Singapore)
Undergraduate Research Program in Statistics, 2023
Diffusion Models for Probabilistic Deconvolution of Galaxy Images
Yuhang Li (Next step: Master’s in Computer Science at University of Illinois, Urbana-Champaigna)
Undergraduate Research Program in Statistics, 2023
Diffusion Models for Probabilistic Deconvolution of Galaxy Images
Zhong Zheng (Next step: Master’s in Computational Data Science at Carnegie Mellon University)
Undergraduate Research Program in Statistics, 2023
Atomic Maps Reconstruction for Cryo-EM Data with Continuous Heterogeneity