CS + cognitive science · Rutgers '28

Aryan
Singh.

I build AI that minds the user — tools that read pages for people who can't, and models that watch roads for drivers who don't.

open to summer 2027 ✦ +33% model accuracy dean's list ×4
morph.exe

Aryan Singh — CS + cognitive science @ Rutgers '28

about the throughline

Every project starts with a person — a reader who can't parse the page, a driver who can't see the crossing. I study how people read, drive, and decide, then build the AI that notices.

education

Rutgers University, New Brunswick — BS Computer Science + BA Cognitive Science, minor in Mathematics. Expected May 2028.

GPA 3.73 · dean's list — 4 consecutive semesters (fall '24 – spring '26)

coursework — data structures · linear algebra & optimization · computer architecture · discrete structures · principles of programming languages · multivariable calculus · intro to AI · distributed systems

toolbox

GoPythonTypeScriptC / C++React / Next.jsNode.jsFastAPISpring BootPostgreSQLRedisKafkaDockerTerraformAWSGitHub ActionsPrometheus · GrafanaRAG · LlamaIndexLangChainClaude APIPyTorchYOLO / CNNs

experience three roles, one throughline roles link to what they shipped →

Nov 2026 – Apr 2027

AI Engineering intern incoming

BalanX Bio remote

  • Designing, testing, and refining AI personalization models.
  • Partnering with the data and engineering teams.
  • Deploying AI features into production.
AI personalizationmodel evaluationproduction ML

June 2026 – Aug 2026

Machine Learning extern

Collaborative Solutions, LLC Piscataway, NJ

  • Led the team's migration from YOLOv5 to YOLOv11.
  • Built the training pipeline across GitHub, Drive, and Colab — 15 experiments.
  • Fine-tuned from COCO weights: +33% mAP50 over the prior benchmark.
  • Found a bug shrinking training images. Thin-object AP +111%.
YOLOv11transfer learningPyTorchColab
shipped Visual distractions — +33% mAP50, 0.406 → 0.606 case study →

Jan 2026 – present

Lead Software Engineer

HCI Lab for Health & Sustainability, Rutgers New Brunswick, NJ

  • Built and deployed a Next.js air-quality platform on weather and OpenAI APIs.
  • Led the 25+ participant study behind it — findings co-authored into a paper submitted to ACM CHI 2027.
  • Merged a Python classifier into ActivityWatch — 19k stars, full PR review.
  • Sole engineer for 15+ researchers: mapping, comparison, streaming chatbot.
Next.jsTypeScriptPythonOpenAI API
shipped What's happening around me — 25+ participant research study repo ↗

projects built on my own time rows open their repo ↗ · case studies open in full →

why I build this way a 20-second demo

Cognitive science taught me the uncomfortable part: people miss what they aren't looking for — even when it's obvious, on screen, in front of them.

Try the demo 20 seconds →

A mock UI flickers. One thing keeps changing. Most people never see it — click through to find out if you will.

off the clock proof I'm a person tap photo editing for the gallery →

soccer

pickup games I take far too seriously

the gym

same discipline as debugging: show up, add one

pokémon

click me — I dare you

photo editing photos →

where I learned that color is a decision

music

headphones on = the code is going well

hiking

best debugging happens away from the laptop