Hi! My name is Ayaan.


I research and build practical AI applications.

As a M.S. in engineering student at Columbia, I pride myself in my problem solving abilities. These days I research data generation in both digital twin and neuroscience applications.

Check out my outstanding student feature, or my DEF CON 33 and 34 presentation.

I love to read, watch cinema, go on long walks, and occasionally blog.

Feel free to connect me on LinkedIn, or view my Google Scholar and GitHub sites. Please email me at: ayaan.qayyum@columbia.edu

Ayaan

Education

Columbia University

M.S. in Electrical Engineering | Sep 2025 — Present

Specialization in Data-Driven Analysis and Computation.

GPA: 3.76/4.0.

Tesla Scholar

Rutgers University

B.S. in Electrical Engineering | May 2025 | Summa Cum Laude

GPA: 3.92/4.0.

Dean's List for all semesters

Professional Experience

  • Designed the first PNA-based GNN that beats four baseline architectures by a wide margin (spearman rho=0.88 versus baseline of 0.59)
  • Derived a first-order Jacobian surrogate for neuron-ablation data labels, a ~1000x reduction at N=1000, with empirical validation
  • Built a 49k parameter model that beats larger models with generalization across network width (N=10 to N=1000)
  • Designed physics-informed synthetic data pipelines for mmWave radar, improving 3D localization accuracy by up to 3.7 AP points
  • Built the paired real-simulated radar dataset of 14 objects, enabling 95.3% 14-way object recognition, co-authored two papers
  • Improved click PR-AUC by 5% for display and sponsored search ads ranking models with a 42% parameter reduction over baseline
  • Invented a SHAP parameter-cost-per-feature metric where PCA failed, flagging 3 defective features, cutting 20% excess model size
  • Diagnosed training instability via per-layer gradient-norm analysis, added residual connections; cutting training epochs by 50%
  • Redesigned the published multi-task hierarchy into a bifocal structure, achieving a 23% param reduction
  • REALTIME team at WINLAB, Rutgers University
  • Built distributed learning algorithms across a 32-node testbed, validating convergence rates in real-world constraints
  • Led a team to conduct large-scale network experiments with Python (Socket, MPI) to evaluate sync/async communication
  • Varied topology, communication frequency and latency to discover breakdown conditions (>14 iterations without update)

On Campus Experience

  • Mentored teams to develop two search-and-rescue drones within 3 ft of localization, two papers at MIT URTC & provisional patent
  • Significantly improved the learning experience in Engineering Economics by introducing innovative teaching methods, such as developing an online interactive learning tool and designing worksheets, used by over 100 students per month
  • Contributed to a 20% increase in the class average scores through effective pedagogical techniques
  • Developed and utilized diverse communication, advising, and teaching skills to enhance student understanding
  • Actively participated in academic success of students, reinforcing key concepts, providing personalized assistance
  • Taught and advised students on Engineering Economics course material, wrote 1000+ textbook questions
  • Taught and engaged 1,000+ students by designing and delivering professional development lectures in class
  • Designed four professional development workshops for classroom use, with lesson plans and teacher training
  • Worked closely with stakeholders to optimize the delivery of intended materials for maximum student impact
  • Webmaster from Aug 2023 to May 2024.
  • Organized and facilitated community-building events and educational programs for residents
  • Provided crisis intervention and support, serving as a resource for student concerns and conflicts
  • Collaborated with residential life staff on policy enforcement and student conduct issues
  • Earned two awards. Helped the organization run conferences, trained fellow volunteers
  • Check out the press release of the ceremony here

Skills

Technologies I know:

  • Python
  • PyTorch
  • TensorFlow
  • Sci-kit Learn
  • PostgreSQL
  • Apache Spark
  • Linux
  • MATLAB
  • Git/GitHub/VCS
  • HTML / JavaScript
  • MPI, Socket, Ansible

Certifications:

  • AWS Cloud Practitioner
  • AWS AI Practitioner

Projects

Unmanned Wireless Penetration Testing Device

Embedded technology platform, fully Kali Linux capable, with ad-hoc LoRa network technology.

Python • Linux • LoRa

News Sentiment Embeddings for Stock Price Forecasting

Explore the benefits of integrating news headline to improve stock price prediction performance.

Python • PyTorch • Pandas

FFT-on-FPGA Learning Tool

Explore learning and coding an FFT in Verilog with a demonstration. Presented at the IEEE ISEC 2025.

Verilog • Python

Password Generator

Generate passwords to use.

HTML • CSS • JavaScript

Rutgers IEEE Micromouse

Our team won 1st place in the regional competition.

HTML • CSS • JavaScript

Morse Code

I made this to learn and practice Morse Code efficiently.

HTML • CSS • JavaScript

QR Code Maker

One professor needed a tool to generate QR codes without ads. So I made one.

HTML • CSS • JavaScript

Calculator for Engineering Economics

As a Learning Assistant, I built a tool to help students. Used by over 100 students per month for the last three years.

HTML • CSS • JavaScript

Word and Character Counter

I catch myself searching up a tool like this very often, so I made one.

HTML • CSS • JavaScript

Tetris

This was a real challenge to build at the time, but I learned a lot and still come back and play with it from time to time.

HTML • CSS • JavaScript