Portfolio

Projects

Projects 1. Build your Own Perplexity

• Perplexity AI is an AI-chatbot-powered research and conversational search engine that answers queries using natural language predictive text. Perplexity is also a search engine. Instead of presenting you with a list of websites that match your query, Perplexity gives you a short summary answer and the references it used to create. In some cases, the summary will be all you need. In others, you'll want to dive into the different sources.

• To build your own Perplexity AI (simplified version), you need only Langchain, an LLM, and a web search API. Here is my own Perplexity; you can try the demo here: (My Own Perplexity demo)here)

Projects 2. Multimodal pdf RAG with langgraph

• LangGraph is a library built on top of LangChain, designed for creating stateful, multi-agent applications with LLMs (large language models). It enables the construction of cyclical graphs, often needed for agent runtimes, and extends the LangChain Expression Language to coordinate multiple chains or actors across multiple steps.

• Pdf files can contain multimodal data, including text, table, figure. Using LangGraph, I built a Multimodal pdf RAG app with web searching function, when you ask something that's not in the pdf file, the app will retrive the answer on internet.

Projects 3. Document Question-Answering bot

• Personalized medicine targets unique group of genes for patient based on the individual's unique genetic background and symptom thus the report building is time-consuming and involoves large amout of documents reading.

• To build a document Question-Answering model use langchain and LLM, this report can be automated, by uploading pdf files to the chatbot, and ask questions. The chatbot will read the pdf files for you and summarize the content. Saved 90% of the time for report building by using the chatbot to read scientific papers for you. (Online pdf chatbot demo)here)

Projects 4. Bitcoin price prediction using Reddit comments sentiment analysis

• Automatically web scraping and sentiment analyze everyday reddit comments regarding Bitcoin.

• Predict next day Bitcoin price using sentiment counts and historical bitcoin price data and daily-updating on a dashboard.

• Web API for sentiment analysis of any input comments about Bitcoin.

Projects 5. Transformers to predict circular RNA

• Adapt transformer NLP model to analyze DNA/RNA sequence.

• Build and deploy a transformer model to predict if any given RNA sequences can form circular RNA, with 0.92 auROC.

• A web App to predict if an input RNA sequence circular RNA.

Projects 6. Zika Virus Drug Design using RNN-LSTM and Proteomics

• Use proteomics analysis to identify proteins associated with Zika virus infection.

• Generate candidate and fine-tune with validated drugs using RNN-LSTM.

• Use virtual molecular docking screening to investigate the binding of drugs to target proteins.

Projects 7 (Bioinfo) Essential Genes from Genome-wide Screening as a Resource for Neurological and Mental Disorders Gene Discovery

• By data mining of 6 publicly available case-control RNA sequencing dataset, we found that NEGs are more likely to exhibit differential expression in brain tissue in neuropsychiatric disorders compared to ACEGs.
• WGCNA analysis of human brainspan dataset, we found that NEGs and ACEGs are enriched in different co-expression networks and show different temporal expression patterns.
• GWAS data and rare variants data for neuropsychiatric disorders shows that ACEGs are exclusively enriched in neurodevelopmental disorders, while NEGs are related to neurodevelopmental, psychiatric, and neurodegenerative diseases.

Projects 8 (Bioinfo) Hepatitis Viruses affect the expression of Endogenous Retrovirus and Tumor Microenvironment in HCC

• single-cell RNA sequencing analysis to deconvolute the immune landscape and tumor heterogeneity in a cohort of patients with hepatitis virus (HBV, HCV, HDV) associated human hepatocellular carcinoma (HCC).
• Re-analyze the bulk RNA-seq data for the cohort to include the expression values of human Endogenous Retrovirus (hERV), and found correlations with hepatitis virus status.

Projects 9 Fatal events in Houston

• Web Scraping fatal events reports from HPD official website, ETL using python and load as data frame.
• Visualization and summarization using Tableau.

Contact

• Author: Wei Zhang
• Email: zwmc@hotmail.com
• Github: https://github.com/vveizhang
• Linkedin: https://www.linkedin.com/in/wei-zhang-76253523/