• Spearheaded the development of critical AI/ML infrastructure systems, including GPFS training job caching solution, TurboCache, Virtual FDS for Spark Data Processing enhancing performance and reliability.

• Drove initiatives to enable encrypted-zone compliance, DMA compliance and optimize large-scale Ads/Feed ML pipelines, improving user experience.

• Innovative 3D Deep Convolutional Neural Network Encoder Decoder system design and employed in generation of ligand surface shape and property distribution.

• Proficient model building and training skill in Python with Keras built on top of Tensorflow.

• Large dataset creation with over 20,000 3D voxelization on point clouds of chemical data and data augmentation with rotation matrix.

• Support Vector Machine and Least Square SVM models built for prediction of radiation absorbance data with r-square of over 0.9 and subjective/objective feature elimination and Y-scrambling method for overfitting reduction

• Using machine learning techniques such as Random Forest, Support Vector Machine to study Quantitative Structure Activity Relationship(QSAR) of chemicals.

• Have done computational biology tasks, such as DNA structure analysis with Hidden Markov Model, DNA motif finding with Gibbs Sampler, DNA consensus sequence finding, phylogenetic trees analysis using R programming language

• Molecular docking for protein design using MOE

• Solid background and knowledge of data structure and algorithms

• Distributed Twitter Service Using the Paxos/Wuu-Bernstein Algorithm with c++ and AWS.

• Content based candidate recommendation system with TFIDF, LDA, LSI on talent profile json data

• Four-pathway Mobile Eye Gaze Estimation with Deep Learning model using Tensorflow.

• Twitter Sentiment Analysis with LSTM Recurrent Neural Networks.

• First author paper accepted by [EMNLP 2018]. Demonstration [link].

• Build innovative system for multiparty meeting dynamics study assistance with Stanford CoreNLP tools and bipartite graph for opinion extraction and monitoring.

• Verbal feature design and calculation for improvement of multimodal meeting processing and prediction and results published in [ICMI2018]

• Participated in LoReHLT 2018 for training low resource language data annotator and system improvement.

• Pipeline design from audio to text transcription using Speech to Text (IBM tool) to data analysis and understanding of multiparty conversation.