- Foundation Models
- Multimodal Forecasting
- Physical Simulation
- Causal Discovery
- Music Generation
- Bio-signal Analysis
- Commodity Prices
- Battery Capacity
- Music Composition
- Molecule Design
- Retail Checkout
- Vision Inspection
- 01인공지능의 핵심 원리를 청중에 맞추어 교양 또는 전공의
수준으로 쉽고 명확하게 설명할 수 있습니다. - 02인공지능과 관련된 배경 지식 (코딩, 수학, 확률/통계) 및
실무에서 얻은 통찰을 전달드릴 수 있습니다. - 03딥러닝의 역사와 최신 연구 흐름 그리고 산업 현장에서의
다양한 적용 사례들을 소개드릴 수 있습니다. - 04실제 산업 사례를 바탕으로 한 코딩 실습과 함께 다양한
주제의 튜토리얼을 직접 진행할 수 있습니다.
Education
* equal contribution
Developing a time-series foundation model for zero-shot forecasting across diverse domains and variates; trained models from 23M to 1B parameters on 2T data points.
Lithium price forecasting with an expert-grade supply-demand-inventory simulator, paired with an agent for advanced error analysis and calibration — spanning domain-expert collaboration and an agent deployment framework.
Extracted causal graphs through continuous data-driven optimization that explain changes in lithium prices, helping human experts make better-informed decisions.
Benchmarked tree-based and neural models for discharge-capacity prediction; proposed a neural model that beats baselines via improved regularization, explained with XAI. Explored life-cycle prediction under data scarcity with battery domain knowledge.
Built a web interface for Chemical VAE that generates novel molecules in SMILES form with specific target properties, helping chemists explore candidate structures.
Applied efficient architectures, pruning, compiling, TensorRT, and Intel MKL on CPU/GPU, reaching 0.03% model size and 0.11% latency on CPU without loss in inspection quality.
Worked end-to-end from data collection and labeling to modeling and deployment; improved the classifier of the object detector for robustness across environments.