cv
Academic CV and selected experience.
Basics
| Name | Doeun Kim |
| Label | M.S. Computer Science Student |
| doeunkim@usc.edu | |
| Github | github.com/doeunyy |
| linkedin.com/in/doeun-kim | |
| Interest | Machine Learning, AI for Healthcare and Biomedicine, Explainable AI, Multimodal Learning, Generative Models |
Education
-
2025.06 - Present Los Angeles, CA
University of Southern California
M.S. in Computer Science
- Expected December 2026
- GPA: 3.88 / 4.0
- Relevant coursework: Machine Learning, Deep Learning, Applied NLP, Information Retrieval & Web Search, Algorithms, AgentAI, Game Development
-
2022.08 - 2022.12 San Jose, CA
San Jose State University
International Student Exchange Program, Computer Science
- Relevant coursework: Introduction to AI, Formal Language and Computability
-
2020.03 - 2025.02 Seoul, South Korea
Ewha Womans University
B.S. in Computer Science and Engineering & B.S. in Content Convergence
- GPA: 4.21 / 4.3 (Top 4%, Summa Cum Laude; Dean's List, all semesters)
- Undergraduate thesis: Layer-Adaptive Quantization of Diffusion Models using Fisher Information
- Relevant coursework: Data Structures, Algorithms, Computer Architecture, Operating Systems, Database, Software Engineering
Research Experience
-
2025.11 - Present Los Angeles, CA
Graduate Research Assistant
University of Southern California
Laboratory for Machine Learning, Health and Biomedicine
Advisor: Prof. Ruishan Liu
- Developed AI-driven deimmunization methods for synthetic binders, identifying high-immunogenicity peptide regions and suggesting amino acid substitutions to reduce predicted immunogenicity while preserving binding affinity.
- Built an interactive research pipeline for HLA-specific peptide scanning and population-aware immunogenicity analysis, using FastAPI, Gradio, RabbitMQ, peptide-level score caching, and beam search to support efficient and scalable evaluation.
-
2024.04 - 2025.02 Seoul, South Korea
Undergraduate Researcher
Ewha Womans University
AI Computing Platform Lab
Advisor: Prof. Jaehyeong Sim
- Led research on Fisher-guided layer-adaptive quantization for diffusion models, reducing model size by 49.4% while improving FID by 7%; presented the work at the 2024 IEIE Symposium as a co-first author and presenter.
- Conducted a DDPM interpretability study using Integrated Gradients, GradSHAP, and Occlusion to analyze the model's generation process across timesteps, showing a progression from global outlines to fine details; presented at the KCC 2024 XAI Workshop as the only undergraduate-led team.
Publications
-
2026 Contrastive Dual Encoders Enable Ultrafast Proteome-Wide Antigen Presentation Prediction across Population-Scale HLA Diversity
Under review
Zeng, B.*, Yang, N.*, Wang, S.*, Choi, J. W., Kim, D., Liu, B., Baker, D., Yu, B.†, & Liu, R.† (*Equal contribution; †Co-corresponding authors)
-
2024.12.27 Layer-Adaptive Quantization on Diffusion Models using Fisher Information
2024 Symposium of the Institute of Electronics and Information Engineers, Seoul, South Korea
Kim, D., Byeon, J., Park, I., & Sim, J. Poster presentation. Co-first author and presenter.
-
2024.06.27 Analysis of Diffusion Model Inference Mechanisms Using XAI Techniques
KCC 2024 XAI Workshop, Jeju, South Korea
Kim, D., Byeon, J., & Park, I. Poster presentation. Co-first author and presenter.
Projects
- 2025.11 - 2026.10
EPIC Research Platform
Full-stack biomedical AI research platform supporting HLA-aware peptide scanning and computationally intensive machine learning inference.
- Independently developed the frontend and backend of a biomedical AI research platform using Gradio and FastAPI, integrating HLA sequence processing, peptide scanning, presentation scoring, and sequence similarity-based retrieval into a unified interface.
- Implemented RabbitMQ-based asynchronous job orchestration so that long-running inference runs on background workers without blocking concurrent API requests.
- Reduced redundant computation through singleton-managed model services, precomputed embeddings, score caching, and persistent result storage.
- 2026.09 - 2026.10
Veridex
Evidence-grounded multi-agent research assistant for querying academic papers.
- Independently developed a supervisor-based multi-agent system using LangGraph with three specialized workers, three custom FastMCP servers, and a retrieval pipeline over 42 papers and 1,855 indexed chunks.
- Designed hybrid OpenAI/Gemini routing with task-specific reasoning budgets, reducing median time to first token from 19.1s to 9.7s and cost per query from \$0.0313 to \$0.0201 with no observed RAGAS degradation.
- Deployed the containerized application to Google Cloud Run with automated CI, reaching RAGAS faithfulness of 0.90 and context precision of 0.96 on a 35-question evaluation set.
- 2026.01 - 2026.05
Multimodal Entrance Detection System
Multimodal geospatial entrance detection using aerial imagery, street-view images, GPS trajectories, and building footprint overlays.
- Led the Kimi-k2.5-based in-context learning framework, designing overlay-grounded prompting that projects building footprints, 64 candidate entrance markers, and pedestrian GPS trajectories onto a shared aerial representation.
- Developed post-hoc confidence calibration using rank-based floors and GPS-density priors, improving F1 from 0.372 to 0.595 and AUPR from 0.175 to 0.450 without task-specific fine-tuning.
- 2025.08 - 2025.12
PokerBench
Small language model training and evaluation for poker decision-making under uncertainty.
- Fine-tuned Qwen3, LLaMA-3.2, and Gemma variants using QLoRA and LoRA adapters on approximately 560K solver-labeled samples.
- Built an evaluation pipeline for instruction-following under uncertainty, comparing model performance against larger proprietary baselines using Action Accuracy and Exact Match.
- 2025.01 - 2025.02
TalesRunner
AI-powered storybook generation from user-provided images.
- Developed a generation pipeline combining BLIP-based image captioning, fine-tuned KoT5 story generation, and GPT-4o-assisted text refinement.
- Integrated captions, story text, and user images into downloadable PDF storybooks, validating narrative quality, image-text alignment, and formatting consistency across 100+ test cases.
- 2024.12 - 2025.02
Jeans
AI-assisted photo editing and sharing service designed for senior users.
- Led a 5-member team in developing AI features including fine-tuned Whisper for Korean dialect speech recognition and YOLO/FaceNet-based facial recognition.
- Deployed FastAPI-based ML inference services on AWS EC2 with GitHub Actions CI/CD, refining speech-driven interaction flows through senior user interviews and usability tests.
Professional Experience
-
2026.10 - Present Santa Monica, CA (Remote)
-
2021.05 - 2022.07 Seoul, South Korea
Software Engineer Intern (Full-stack)
RainbirdGEO
- Developed full-stack features for a weather-forecasting mobile app using React Native and Node.js/Express, supporting pilot deployment and usability testing in Southeast Asian regions.
- Conducted product analysis using user feedback, competitor research, and 2,000+ survey responses, creating Tableau dashboards and contributing to improvements in satisfaction scores and forecast accuracy.
Teaching & Mentoring
-
2025.09 - Present Viterbi Graduate Mentor / Women in Engineering Mentor
USC Viterbi School of Engineering
Mentoring incoming and women graduate students on academic planning, research involvement, and transition to graduate student life.
-
2023.09 - 2025.05 Instructional Assistant (Part-time)
Scholar English - Private English Academy
Supported instructional operations across grading, curriculum revision, and student communication, while developing a custom GPT-based chatbot to improve workflow efficiency.
-
2021.03 - 2021.12 Volunteer Tutor
Neulhaerang School - Inter-University Volunteering Org. (Ewha Womans Univ. & SNU)
Taught weekly computer literacy classes for elementary students and created accessible learning materials for beginner learners.
Awards
- 2025
Individual Best Award
SK Telecom FLY AI Program
Awarded by SK Telecom FLY AI Program, run by South Korea's largest telecom company.
- 2025
Outstanding Project Award
SK Telecom FLY AI Program
Project award from SK Telecom FLY AI Program.
- 2022-2024
National Scholarship for Science and Engineering
Korea Student Aid Foundation
Full tuition merit-based scholarship for two years.
- 2024
Capstone Project Merit Award
Ewha Womans University
Awarded at Ewha Capstone Design Poster Session.
- 2020-2024
Honors Scholarship
Ewha Womans University
Awarded four times.
- 2022
International Study Scholarship
Ewha Womans University
Scholarship supporting international study.
Skills
| Programming | |
| Python | |
| JavaScript / TypeScript | |
| SQL |
| AI / ML | |
| PyTorch | |
| Hugging Face Transformers | |
| scikit-learn | |
| LoRA / QLoRA | |
| LLM & SLM fine-tuning | |
| Vision-Language Models | |
| Model evaluation |
| LLM Systems & Agents | |
| LangGraph | |
| LangChain | |
| MCP / FastMCP | |
| Retrieval-Augmented Generation | |
| Tool calling | |
| Prompt engineering | |
| RAGAS |
| Backend & Web | |
| FastAPI | |
| Django | |
| Node.js / Express | |
| React Native | |
| Gradio | |
| RabbitMQ | |
| REST APIs | |
| Server-Sent Events |
| Cloud & DevOps | |
| AWS (EC2, Lightsail, Elastic Beanstalk) | |
| Google Cloud Run | |
| Docker / Docker Compose | |
| NGINX | |
| GitHub Actions CI/CD | |
| pytest |
| Data & Storage | |
| MongoDB | |
| MySQL | |
| SQLite | |
| Firestore | |
| Chroma | |
| NumPy / Pandas | |
| Tableau |