Product, engineering and researchNew York, NY

Ayo
Adetayo

Studying computer science at Columbia University. Recently in product engineering at SpaceX; formerly in product leadership at Amazon.

View profile
Ayo Adetayo
01 / Experience

Professional experience

SpaceX2026

Technical Product Engineer Intern

May–Oct 2026
Starlink Enterprise Product Architecture

Led product strategy and customer discovery for AI-cluster networking, translating workload requirements into network topology, capacity, and service-level requirements. Defined pricing and service tiers for dedicated internet access, and led engineering delivery of a unified Starlink business ordering system across service plans and hardware.

Amazon2019–2026

Senior Product Manager

Jan 2025–Feb 2026
Amazon Business

Product selection, probabilistic demand forecasting, inventory placement, and replenishment recommendation systems for business customers.

Senior Policy Product Manager

Feb 2024–Jan 2025
Amazon Creators

Session-level attribution, personalized recommendations, and machine-learning systems for content evaluation and platform integrity.

Continuous Improvement Manager

Jan 2023–Feb 2024
Last Mile Operations

Routing optimization and delivery-quality analysis using geospatial and delivery-density data across Los Angeles and San Diego.

Program Operations Management

Mar 2019–Jan 2023
Program Operations

Cross-site process improvement, program deployment, and operational standardization across Amazon’s fulfillment network.

  • Operations Program Manager III, Operations, L6 05/2022–01/2023
  • Area Program Manager II, Operations, L5 07/2021–05/2022
  • Area Program Manager I, Operations, L4 10/2020–07/2021
  • Shift Assistant Operations Manager, L3 03/2020–10/2020
  • On The Road Operations Team Lead, L2 01/2020–03/2020
  • Learning Ambassador, L1 03/2019–01/2020
02 / Research

Research and projects

Explore projects
Independent research · 2026

Comparing CNNs and Vision Transformers under matched resource budgets on CIFAR-100

A controlled CIFAR-100 comparison of a residual CNN and a Vision Transformer trained from scratch under matched parameter and estimated compute budgets, measuring accuracy, calibration, noise response, and runtime to inform model selection.

Read the study
Top-1 accuracy · 0–100%Higher is better
CNN60.96%
ViT55.62%

Higher CNN accuracy on clean images

The CNN achieves a 5.34 percentage-point accuracy advantage under the shared training protocol.

Custom models with approximately 1M parameters · CIFAR-100 · Random initialization · Three-seed means

Columbia · Coauthored report · 2025

Dimensionality Reduction Techniques: A Comparison of Principal Component Analysis and Linear Autoencoders on the MNIST Dataset

Three MNIST experiments compare reconstruction and subspace geometry, then test architectural constraints and training choices. Includes the full paper and all four original figures.

Read the research
Original MNIST digits and their PCA and linear autoencoder reconstructions from the paper
Original digits, PCA and linear autoencoder reconstructions · 64 latent dimensions
03 / Writing

Academic and cultural essays

View writing
Constitutional history

Constitutionally Bankrupt in the Marketplace of Ideas

The limits of First Amendment protection for college student protest and speech.

Constitutional history

The Sufficient and Necessary Conditions to Govern a Union

Marshall, Taney, federalism, slavery, and the legacy of Dred Scott.

Politics and institutions

Society, State and Democratic Erosion

Executive aggrandizement and the changing relationship between state power and societal constraint.

Culture and economics

Racial Capital and the Blueswomen’s Voice

Analysis of Ma Rainey, Big Mama Thornton, and the political economy of artistic recognition.

04 / Personal

Beyond work

More about me

I love hanging out with friends, trying new restaurants in New York, traveling, and taking road trips. I recently drove across the country from New York to Los Angeles twice. I’m also a big music lover.