PhD Researcher · Computer Science · The Ohio State University

Building machines that listen.

I build deep-learning systems for speech, sub-kilobit neural audio coding, and decoding auditory attention from EEG — end to end, from custom electrodes to model design to deployment.

See the work Download CV

Advised by Dr. Donald S. Williamson · Publishing as S. A. Alavi Bajestan · Columbus, Ohio

fs 16 kHz< 1 kbpsEEG · BCI

What I work on

  1. 01Speech processing
  2. 02Neural audio coding
  3. 03Auditory attention decoding · EEG
  4. 04EEG-informed speech enhancement
  5. 05Brain–computer interfaces
  6. 06Foundation models
  7. 07Reinforcement learning for LLMs

Selected work

Research & systems I built, most recent first. In submission & preprints.

SubmittedInterspeech 2026Lead

CADENZA

Content-adaptive neural audio codec at sub-kilobit rates.

FSQ · Flow matching · Streaming Conformer · <1 kbps

SubmittedNeurIPS 2026

NEUROTOKEN

Joint source / direction attention decoding + envelope from short EEG windows.

EEG · AAD · Conditional flow matching

SubmittingIEEE TASLP

MAESTRO

Multimodal dataset — EEG + eye-tracking + body — for auditory attention.

Dataset · Eye-tracking · Multimodal

TargetingNeurIPS 2026

CorticalFlow

EEG-to-audio decoding via envelope-first multi-resolution reconstruction.

EEG→Audio · Multi-resolution · Envelope

ResearchContributor

AIMO3

Reinforcement learning (GRPO / RLVR) for math-reasoning LLMs, with MCTS and Lean 4 verification.

GRPO · RLVR · MCTS · Lean 4

PreprintarXiv

Contrastive-AAD

Contrastive-learning attention detection on a custom high-density EEG + gaze rig.

Contrastive · IRB rig · HW sync

How I work

I build the whole stack — not just notebooks. Custom high-density EEG + gaze hardware, model design, and deployment.
  1. 01Data + Hardware
  2. 02Model Design
  3. 03Deployment

Custom high-density EEG + gaze + behavioral rig · IRB-approved · hardware-synchronized.

Experience

  1. 2025

    Computer Vision Engineer — BNtelligence

    End-to-end defect detection: data, classification & segmentation, and a real-time video-inference API in production.

  2. ’19–20

    Deep-Learning Consultant — AI-bridge, Hamburg

    Shipped production DL — image enhancement, body-part segmentation, super-resolution (TF / PyTorch).

  3. 2019

    Software Engineer — Payafanavaran

    C# app for 3D surface reconstruction from structured-light scanning.

  4. 2016

    Hardware Design Engineer — Tarashe Sanat

    High-speed FPGA board — DDR3, Ethernet, peripherals — for industrial data acquisition.

Education

  1. 2022 →

    The Ohio State University Current

    Ph.D., Computer Science · adviser Dr. Donald S. Williamson.

  2. 2022

    Indiana University

    Ph.D., Computer Science (transferred to OSU).

  3. ’16–20

    University of Tehran

    M.Sc., Communication Engineering.

  4. ’12–16

    Ferdowsi University of Mashhad

    B.Sc., Electrical & Electronics Engineering.

Tools I work in

Languages

PythonC/C++C#CUDAMATLABJavaJavaScript

ML frameworks

PyTorchTensorFlowLightningscikit-learn

Infra / MLOps

GitLinuxDockerHPC / SlurmWeights & BiasesHydraLaTeXLean 4

Hardware

FPGA · VHDL/VerilogARM Cortex-MLabVIEWCadence

Honors & service

  • CCBS Summer Research Award — The Ohio State University2023
  • 3rd place — CMDC Hackathon—
  • Reviewer — EACL AfricaNLP Workshop2021
  • Student Volunteer — ICLR & ICML2021

Let’s build systems
that listen.

Email · OSUalavibajestan.1@osu.edu Email · Personalsa.alavi.b@gmail.com RésuméDownload CV (PDF) ↓ CodeGitHub (opens in new tab) ↗ PublicationsGoogle Scholar (opens in new tab) ↗