Portrait of Erik Kizior

Erik Kizior

Software & hardware engineer · M.S. EE at Stanford

Stanford, CA

About

I'm an electrical engineering master's student at Stanford who works across the stack, from agentic AI systems down to RTL. This summer at Meta I built an AI platform that triages product signals for Instagram. The summer before, at Amazon, I built a generative-AI engine that automates security threat reviews. At Stanford's Tambe Lab I research hardware for machine learning. At Berkeley I spent two years teaching data structures, most recently leading a 48-person staff for an 800+ student course. I care about systems that are measurable, reliable, and easy for the next engineer to pick up.

Experience

  • Graduate Researcher · Stanford University, Tambe Lab

    –Present

    Computer architecture research on hardware systems for machine learning.

  • Software Engineer Intern · Meta, Instagram Metrics

    Designed and shipped an agentic AI platform (Hack, React) that classifies UI-detected signals into Instagram's metric catalog. It reached ~14% higher coverage than human triage at ~98% precision and saves an estimated 152 engineer-hours per year.

  • Software Development Engineer Intern · Amazon, Application Security

    Built a generative-AI threat analysis engine on AWS (Bedrock, Lambda, Step Functions) that uses a two-stage RAG pipeline to read design diagrams and flag security threats. It supports 20+ security engineers and saves ~120 engineer-hours per year.

  • Undergraduate Researcher · UC Berkeley, MRI Reconstruction

    –

    Trained a PyTorch U-Net to reconstruct MRI scans from undersampled k-space data (fastMRI, 56k+ samples), reaching >0.95 SSIM. Cut training time by more than 10x with distributed GPU training on the Savio HPC cluster.

  • Head TA & Instructor · UC Berkeley EECS, CS61B/CS61BL

    –

    Co-taught the summer course to 350+ students with 96.7% satisfaction, then led a 48-person teaching staff for an 800+ student course. Built autograders and internal course tooling along the way.

  • Tutor Enrichment Director · The Music Connection

    –

    Recruited 50+ tutors per semester for free K–12 music lessons in Berkeley and Oakland. Built a Gmail API tool that automated tutor–family matching and raised pairings by 37.5%.

More Projects

  • A full RTL-to-GDS 4×4 systolic array for ResNet-18 on SkyWater 130nm, later re-implemented in Catapult HLS and scaled to 16×16.

    • Verilog
    • C++ HLS
    • Synopsys
    • Cadence Innovus
    • SKY130
  • Pipelined RISC-V CPU

    A 3-stage RV32I processor on FPGA whose forwarding logic resolves every read-after-write hazard without stalling.

    • Verilog
    • RISC-V
    • FPGA
  • Secure File Sharing System

    End-to-end encrypted file storage and sharing using AES, RSA, and SHA-256, tested against man-in-the-middle and replay attacks.

    • Go
    • Cryptography
  • Git from Scratch

    A version-control system with commits, branching, and merge conflict resolution, built on SHA-1 content-addressed snapshots.

    • Java
    • File Systems
  • Pacman AI Agents

    A* search, minimax with alpha-beta pruning, Q-learning, and HMM-based ghost tracking from noisy sensor data.

    • Python
    • Reinforcement Learning
    • Probabilistic Inference
  • Voice-Controlled Car

    A custom mic board with a bandpass filter that feeds PCA and k-means speech classification to drive an Arduino car.

    • Arduino
    • Python
    • Signal Processing

Education

  • M.S. Electrical Engineering · Stanford University

    –
  • B.S. Electrical Engineering & Computer Science · UC Berkeley

    –

    Cum Laude · Tau Beta Pi · Eta Kappa Nu · Outstanding GSI Award