Hi, I’m Nathan Samson

Senior at UMBC

I find it really cool to build, but I also really care about privacy and safety.

This is the earliest project I can find of me coding when I was 8.

Click this if you want to learn more.

Here’s some of my experience.

Publications

  1. C. Badolato, N. Samson, H. Hajj Hassan, C.-K. Huang, G. Bouloukakis, P. Pappachan, R. Yus.

    PSMark: A Distributed IoT Benchmark for Publish/Subscribe Under Domain-Based Workloads.

    IEEE International Conference on Pervasive Computing and Communications (PerCom), 2026. Artifact Certified. Paper

  2. N. Samson.

    Do Robot Safety Judges Confuse Success with Safety?

    Science of Physical AI Safety (SPAIS) Workshop, Conference on Robot Learning (CoRL), 2026. Accepted. Code

Experience Research Experience

Experience

Grassroots Analytics

Jan 2026 – present

Junior AI Engineer · Aug 2026 – present

Previously AI Engineer Intern · Jan – Aug 2026

  • Built a news pipeline that ranks relevant coverage and an article library for staff research and AI-drafted content.
  • Built chat for news search and message drafting. Now building tools to group donors and predict who is likely to give.

Capital One

Jun – Aug 2025

Software Engineer Intern

  • Diagnosed a timeout that silently broke production API monitoring and helped migrate the service from AWS Lambda to Fargate.
  • Added logs, alerts, incident reports, and a results dashboard so engineers could catch failures and investigate them.

OmniSyncAI

May – Jul 2024

Software Engineer Intern

  • Built company onboarding from scratch in React, Node.js, and PostgreSQL, including account setup, authentication, and team invitations.
  • Built an AI recommendation engine that suggests teammates and configurations based on each company’s signup needs.

DAMS Lab, UMBC

Sep 2023 – present

Undergraduate Researcher

  • Built a sensor pipeline and React dashboard for a campus system that estimates and forecasts room occupancy from laser scanner readings.
  • Developing SPACES course labs that teach students to connect sensors, exchange messages, and build data pipelines.

Research experience

Durable Guarantees

First author

  • An audit of whether sensitive attributes remain recoverable after a model claims to remove them.
  • Tests representations and model outputs with stronger attackers, then measures the accuracy cost of preventing leakage.
Code

PCRL

First author

  • A representation learning method that produces a different view of the same data for each intended use.
  • Trains a shared encoder with privacy constraints for each purpose, with audits for leakage of sensitive attributes.
Code

MQTT-DAP

Co-author

  • A framework for enforcing consent and permitted data uses within the MQTT protocol.
  • Supports access, correction, and deletion requests, with a benchmark for privacy correctness and performance.

Submitted to ACM SenSys 2027

BrokerBenchmark

PSMark

Co-author

  • A distributed benchmark for comparing publish/subscribe systems under realistic IoT workloads.
  • Models varied devices and behaviors from real datasets, enabling consistent comparisons across MQTT and DDS.

Published at IEEE PerCom 2026 · Artifact Certified

PaperCode