Aleksey Valouev

Aleksey Valouev

EECS, University of California, Berkeley
Robotic AI and Learning Lab (RAIL)

EECS undergraduate at UC Berkeley. Research on robot learning and computer vision. Working on fast and effective embodied memory for navigational VLA models. Advised by Catherine Glossop and Sergey Levine.

valouev@berkeley.edu LinkedIn GitHub

Education

University of California, Berkeley

GPA: 4.0 / 4.0

B.S. in Electrical Engineering and Computer Sciences

Relevant coursework: Operating Systems (A+), Machine Learning (A), Probability (A), Data Structures (A), Computer Architecture (A), Structure and Interpretation of Computer Programs (A)

Teaching

Undergraduate Course Staff 1, CS 162: Operating Systems

  • Selected to support instruction in concurrency, synchronization, virtual memory, file systems, and distributed systems

Publications

  1. Implicitly Learned Neural Phase Functions for PSF Engineering. Presented at ICVISP. arXiv:2410.05413.
  2. [WORKING TITLE: Embodied Memory for Navigational VLA Models] Targeting publication in ICLR 2027.

Experience

Robotic AI and Learning Lab (RAIL)

  • Developed a flexible, multimodal, task-conditioned memory for long-horizon embodied navigation and VLA steering

Darrell Computer Vision Group

  • Built a multi-view vision transformer with temporal cross-attention for fine-grained object detection with PyTorch
  • Deployed on 1 million satellite images for humanitarian debris analysis, increasing mean F1 score from 52% to 65%
  • Composed weekly presentations of key findings and research directions for faculty and industry partners at DARPA

Boston University Computational Imaging Systems Lab

  • Collaborated with a team of 4 to integrate neural networks into a simulator for computational optics and imaging
  • Implemented neural-network based optical focus engineering for high-resolution extended depth microscopy
  • Achieved a median MSSIM of 0.816 and PSNR of 10.38 dB versus 0.000 and 6.65 dB for pixel-wise optimization
  • Authored and presented a paper at ICVISP: Implicitly Learned Neural Phase Functions for PSF Engineering

University of Southern California, Marshall School of Business

  • Built Python, NLP, and web scraping pipelines to analyze thousands of posts from hundreds of short-seller accounts
  • Prepared a report on language-based clustering of faculty research and presented findings at USC CETAFE conference

Chrome App Developer

  • Developed and managed a Google Chrome extension to help students manage their grades in a portable manner
  • Spearheaded expansion to 40 different countries, with over 4,000 installs and over 1,000 regular weekly users

Palo Alto Robotics

  • Oversaw a team of 50+ students in the design, construction, and programming of a FIRST Robotics (FRC) robot
  • Spearheaded competitive preparation and strategy, executing a year-long plan to recruit and train new members
  • Guided the software subteam to the world championship for the first time since 2021 with a technical overhaul

Selected projects & awards

Additional information

Programming languages
Python, Java, C, C++, Bash
Machine learning
PyTorch, JAX, scikit-learn, transformers, computer vision, NLP, object detection
Systems and tools
Linux, Git, Docker, Slurm, PyTorch DDP, ROS, Isaac Sim, OpenCV, Pandas, NumPy, GIS