Web | Linkedin | GitHub | GitLab | Gists
Work Experiences
- Computational Protein Engineer / Machine Learning
Aether Biomachines, Menlo Park, CA
- Startup! Data Science + Machine Learning + Computational Chemistry (Enzyme:Substrate) + MD Simulation
- Postdoctoral Fellow (Computational Simulation/Modeling)
Genomics Institute of the Novartis Research Foundation, San Diego, CA
- Implemented Monte Carlo based structure-prediction algorithm from scratch in Python by referring to research articles and patent documents.
- Totally redesign the above algorithm by accelerating data-generation (GPU-enabled simulation) and incorporating statistical metrics into more generic and robust growth logistic model in Python, R and SQLite.
- Developed and deployed the integrated bioinformatics tool for internal service for reconstructing antibody sequence by leveraging combined Databases and Dynamic Programming.
- Postdoctoral Fellow (Computational Simulation/Modeling)
Beckman Research Institute at City of Hope, Duarte, CA
- Optimized the high-dimensional biological sampling space by refinement protocol, which was enhanced by reusing the robotics locomotion algorithm used in NASA/JPL.
- Created an automation workflow in Python and R for systematic comparison of sampling performance by uti- lizing dimensionality reduction (PCA).
Education
- MS. Computer Science (3.95/4.0), San Jose State University, CA
- Relevant CS Graduate Coursework: Design/Analysis of Algorithms, Deep Learning for Computer Vision, Machine Learning in Security, and Mining of Massive Dataset)
- PhD. Computational Chemical Physics, The Ohio State University, OH
- BS. Chemistry & Physics, SookMyung Women’s University, Seoul, Korea
Technical Project Experience
- [AI] Multi-Agent Deep Reinforcement Learning (DRL) for Walker Systems —
- Solved the challenging problem of robot’s walking and carrying a package in high-dimensional, continuous state/action space using collaborative training with multiple agents and tuning the optimal hyperparameter set.
- [NLP] Semantic Textual Similarity (STS) using Transfer Learning and Embeddings —
- Built a semantic search engine as a pragmatic application of STS in PyTorch/Python with a faster similarity search by approximate k-NN.
- [NLP] Sentiment Analysis using Machine Learning and Deep Learning —
- Comparative study on NLP techniques in capturing the simple polarity of lecture evaluation reviews : from primitive metric (TF-IDF, 80% accuracy), static embedding (Word2Vec; 65%) to state-of-art dynamic embedding (BERT model, 75%)
- [DL/Vision] Lighter, Faster Semantic Segmentation by Post-Training Quantization and Quantization-Aware Training —
- Systematically tested and analyzed performance of quantization as a model compression in conjunction to different fine-tuning protocols for the MobileNet-based segmentation framework.
Computer Skills
- Programming Languages—[Advanced] Python, R and Bash Script; [Skilled] Java, C and Verilog with ModelSim
- Machine Learning/Deep Learning/NLP—Scikit-learn, Pandas, Jupyter Notebook, Tensorflow, PyTorch and BERT
- Deep Reinforcement Learning—Ray RLlib (both single-agent and multi-agent), OpenAI Gym
- Computing Environment—Linux (Ubuntu), HPC (MPI), GPU Computing, AWS (Cloud), Anaconda (Virtual Env)
- Full Stack Web—[Front] HTML, CSS, Bootstrap, Javascript/jQuery, Ajax; [Server] JSP/Tomcat; [Back] MySQL