Ashwin Mahadevan

Experience

  1. March 2023 - May 2025

    Founder and CEO

    PickOne.ai Berkeley, California

    • Built a full-stack AI chat and image application serving thousands of users daily.
    • Designed a semantic recommendation engine to predict user image preferences in a few-shot environment, outperforming recommendations based on similarity metrics.
    • Trained image (diffusion) and text (transformer) models with reinforcement learning on proprietary user preference data collected by the application.
    • Designed a self-hosted server to scale multiple production ML models on a single machine.
    • Machine Learning
    • Generative Models
    • Reinforcement Learning
    • Full-Stack Development

  2. May 2023 - July 2023

    Force Projection Engineering Intern

    Johns Hopkins University Applied Physics Laboratory Laurel, Maryland

    • Optimized the latency of real-time guidance, navigation, control, and targeting algorithms used in US Navy submarines.
    • Analyzed the performance of Kalman-filter-based sensing approaches in real-world environments that violate the Gaussian-distributed errors assumption.
    • Developed a MatLab tool to clean and visualize submarine data from internal tests and NATO wargaming, enabling streamlined analysis by Navy personnel.
    • Obtained and maintained a United States Department of Defense SECRET level clearance.
    • Signal Processing
    • Feedback Control Systems
    • Algorithmic Optimization
    • Operational Security

  3. September 2022 - May 2023

    Zero-Knowledge Proofs Research Assistant

    Simons Institute for the Theory of Computing Berkeley, California

    • Repurposed the MAGE MPC memory preprocessor for use with the TurboIKOS protocol.
    • Optimized the TurboIKOS non-interactive ZKP protocol for use in text-processing.
    • Investigated the feasibility of the TurboIKOS protocol for ZKPs in the legal field.
    • Cryptography
    • Zero-Knowledge Proofs
    • Compilers
    • CPU Memory Architecture

  4. September 2021 - May 2022

    Financial Research Assistant

    Haas School of Business Berkeley, California

    • Scraped, cleaned, and merged cryptocurrency historical data from dozens of exchanges.
    • Modeled price movements in response to social media posts by cryptocurrency influencers.
    • Created event-study plots to visualize orderbook dynamics before and after specific posts.
    • Data Scraping Cleaning
    • Financial Modelling
    • Data Visualization and Analysis

Education

  1. Class of 2025

    University of California, Berkeley

    Applied Mathematics + Computer Science (Double Major) Berkeley, California

    • Competed for the Mens’ Lightweight Rowing team, attending daily 4:30AM on-water practices supplemented by weightlifting and strict dieting.
    • Volunteered for the Computer Science Mentors association, teaching seminars on misunderstood topics in introductory computer science courses.
    • Took a break after my junior year to found PickOne.ai