Interest
Develop innovative ML models for cutting edge biomedical applications:
I am an experienced machine learning data scientist with a PhD background in Computational Neuroscience. My project interface various machine learning domains, including dynamical systems, LLMs and GNN, serving the aim to facilitate mechanistically backed target nominations and perturbation simulation. I am a frequent participant in medtech Hackathons, where I enjoyed creating novel analytical solutions for complex real-world problems. My career goal is to bring disruptive impact to our society via cutting edge technologies.
Current Position
Data Scientist - Omics data - Machine learning application, GSK
Develop ML pipeline to perform target nomination, causal network inference and in silico perturbation
- Design and establish workflow that integrate different ML tools and multi-omics readouts for target nomination, in silico perturbation and mechanistic inference
Benchmarking and optimization of statistical, LLMs and generative models
- Dynamical systems, foundation models, VAE, GNN, LLMs evaluation of predictive output
Work experience
Research Associate , Imperial College London
Annotation of patient disease profile via natural language processing of clinical documents
- Optical character recognition, RegEx pattern matching, Name Entity Recognition,pre-trained Bert text embedding
Genome wide association study of neurodegenerative diseases
- Detect potential causal genetic variants in a sample that comprised of 2.6 millions Single Nucleotide Polymorphisms (Linear Mixed Effect Model, LDSC, MAGMA, gene set enrichment)
- Evaluate the causal relationship of hit genes using eQTL data (Bayes Factor colocalization), single cell RNAseq (MAGMA cell type specificity), and tissue transcriptomics (WGCNA)
Designed a browser application for users to search for genotypes information (R shiny)
Past Project
Principal Investigator, The Alan Turing Institute Data Science Hackathon
• Modelling Amyloid Beta Plaque Formation in Alzheimer's Disease using VAE and VGG-16
- Development of image segmentation methodology to automatically extract amyloid plaque regions from immunohistochemistry stainned brain images
- Leverage Variational Autoencoder (VAE) and pretrained VGG-16 convolutional neural network for the extraction of amyloid plaque morphological features
Work experience
Research Associate , University College London
• Identify the key features in brain oscillation that influence motor-behavioral outcome
- Time frequency spectral analysis, linear mixed effects models, permutation statistics
- Created recurrent neural networks (LSTM, GRU) and Bayesian optimized XGBoost models to
learn the relationship between brain oscillations and motor behavior
- Using SHAP approach to evaluate the key determinants in model decision
• Generated a data processing pipeline for the neurophysiological and behavioral data
- Multi-sources data extraction, alignment, noise detection and removal
- Neuronal spike sorting: filtering, SVD whitening, PCA, superparamagnetic clustering
• Designed and engineered training devices for monkeys
- Programmed an android app and Arduino device for cue delivery and trial counting
• 3D registration of brain recording location using stereotaxic and MRI image coordinates
- MRI brain image processing, 3D rigid transform
• Performed brain implants and electromyography surgeries as the second surgeon
Education
Computational Neuroscience
• Grade: “magna cum laude”
• Investigation of a novel mechanism in synaptic vesicle release
- patch clamp recordings, virus production and purification, shRNA design, cloning
• Validation of a mathematical model of calcium dynamics in synaptic boutons
- Optical device assembly and imaging parameters optimization
- Designed a high-throughput pipeline for quantitative time serious image analysis
- image alignment, noise filtering, artifact removal, automatic ROI detection
• Supervision of undergraduate and master students
• Teaching of 10-15 undergraduate students at practical laboratory sessions
• Organization of conferences
Neuroscience
• Investigation of the effects of iron-loading on the memory formation in the rat
- Multi-electrode field potential recording of hippocampal long term potentiation
- Acute brain slice preparation, transcardial perfusion, brain region dissection
• Teaching of 40-60 medical students at practical laboratory sessions
Molecular Biotechnology
• Genetic association study of Reading Disability
- TaqMan Single Nucleotide Polymorphisms genotyping
Measurement of mRNA expression levels in human tissues