Northwestern University · CS & Violin Performance

Ryan Chun

I build AI systems for signals and structure, from power-grid risk models to violin intonation analysis, and I ship them to the cloud.

View work GitHub LinkedIn
GPA
3.95
Class of
2028
Focus
AI / ML · Audio · Cloud

01 Experience

AI Engineering Intern

ICF International

Jun – Aug 2026
  • Threshold calibration. Recalibrated GridSFM's PGLIB-trained detection threshold (1.06 → 0.89) against ERCOT ground-truth data, lifting near-miss F1 from 12.5% to 90.3% on the 6,852-branch Texas transmission network.
  • Full-stack demo. Built a React + FastAPI app, containerized with Docker and deployed to Azure Container Apps: real-time GNN risk map (3,889 buses), MATPOWER import pipeline, 12-month CMIP6 climate sweep UI, and a GPT-4.1 query interface via Azure AI Foundry.
  • Cloud deployment. Architected CI/CD with Azure Container Registry, Container Apps, and OIDC Workload Identity Federation for keyless GitHub Actions deploys, with digest-based rollout for deterministic rollback.
  • PyTorch Geometric
  • GNN
  • React
  • FastAPI
  • Docker
  • Azure

Research Intern, AI in Music

Purdue University

Jul – Sep 2024
  • Developed audio processing software integrating ML and DSP algorithms to improve editing precision in music production workflows.
  • Built and iterated on an NLP-driven voice-command interface for sound-editing automation.
  • Python
  • DSP
  • NLP

02 Projects

AuditionMate system architecture diagram: audio and score inputs flow through DTW alignment, onset detection, F0 estimation and harmonic NMF into an intonation summary and Claude API feedback.
Creator & Lead Developer · Jan 2025 – Present

AuditionMate

An AI violin performance analyzer that turns a recording and a score into teacher-grade feedback. The pipeline runs CQT chroma extraction, score-informed DTW alignment, pyin F0 estimation and score-informed NMF, then translates the metrics into natural-language coaching with the Claude API.

  • Spectral-flux onset detection plus adaptive passage classification raised measurable note coverage from ~17% to ~70% on dense repertoire (Paganini Caprice No. 1).
  • Structured as an 8-module Python package with a CLI and a yt-dlp ingestion pipeline for YouTube recordings.
  • Python
  • librosa
  • music21
  • DTW
  • NMF
  • Claude API
Explore architecture → Source

03 About

Education

Northwestern University

B.S. Computer Science & B.M. Violin Performance · GPA 3.95 · Expected May 2028

Coursework: Machine Learning, Data Structures, C++, CS Fundamentals, Linear Algebra, Differential Equations, Design Thinking & Communication

Skills

Languages
Python, Java, C++
AI / ML
PyTorch, TensorFlow, PyTorch Geometric, GNNs, LLM prompt engineering, onset detection, DTW
Tools
Docker, Azure (Container Apps, ACR, OIDC), FastAPI, React/Vite, librosa, scipy, music21, ffmpeg, Git

Leadership

  • Co-Founder & Co-President, Python Pals (2021 – 2025): peer-learning program teaching Python and Java to 50+ students
  • Vice President, McLean Highlander Math League (2023 – 2025)
  • Concertmaster, American Youth Philharmonic Orchestra (2023 – 2024)

Honors

  • U.S. Army Orchestra Young Artist Competition, 1st Prize (2024)
  • GMU Calculus Olympiad, 1st Place Team (2024)
  • AIME Qualifier (2024)

04 Contact

Open to software engineering and AI/ML internships.

ryanchun2029@u.northwestern.edu

LinkedIn GitHub