Developed advanced cryptographic protocols using Rust and Halo2 to implement zero-knowledge proofs (ZKPs) for JWT verification. This included setting up cryptographic parameters, generating SNARKs, and aggregating proofs to enhance verification efficiency and security. Leveraged libraries like snark-verifier-sdk and Arkworks to ensure secure cryptographic operations. Responsibilities included designing circuits for JWT validation, managing proof generation and aggregation, and optimizing performance. Integrated logging and benchmarking tools to improve efficiency and reduce resource usage. Implemented error handling and debugging to maintain system integrity.Built a webcrawler using Python and Selenium to extract Twitter comments, handling authentication, session management, and dynamic content loading. Cleaned and processed extracted data by removing URLs, duplicates, and nonsensical entries. Utilized GoogleTranslator API for translating multilingual content.Conducted sentiment analysis using TextBlob, classifying comments as positive or negative. Applied CountVectorizer to identify key topics and used OpenAI's GPT-4 via LangChain for topic summaries. Visualized data with WordCloud and Matplotlib, and automated the entire pipeline for data extraction, analysis, and reporting.

Developed an AI-powered debt collection platform capable of sustaining multiple rounds of conversations and autonomously calling users with a particular emphasis on integrating external databases to transform LLM into a specialized debt collectionautomationCurated a focused dataset comprising 20 debt collection dialogue scenarios, leveraging advanced prompt techniques and utilized iterative software testing and debugging to overcome issues of irrelevant responses and assessed the cost-effectiveness of few-shot learningDirected development strategy for a specialized debt collection language model; analyzed GPT dialogues, identified domain challenges, and emphasized integration of repayment methods and customer service insights to boost memory, relevance, and response accuracySpearheaded a patent application regarding a payment method including user terminals integrated withblockchain and customization smart contractingConceptualized and executed a secure username-password interface for a smart contract website, enhancing user experience while ensuring stringent security protocols

Gathered data on lumber prices, interest rates, inflation rates, cyclical unemployment rates, and other relevant factors influencing the housing market from FRED databaseUtilized Python for thorough data cleaning, including addressing missing data, optimizing data points, refining data fields, and identifying and handling outliers that could impact model accuracyEmployed correlation analysis and business acumen to select pertinent features, developed a regression model for predicting housing market fluctuations with a high accuracy rate of over 90%, and provided valuable insights into the key factors driving the model's predictionsPresented concise summaries of key findings, along with the methodology employed, to the leadership team, incorporating visually appealing visualizations through an intuitive PowerPoint presentation

I accompanied the director of sales to potential client meetings and prepared materials, and also kept itineraries for other sales employees. Furthermore, I was tasked with translations and communication during meetings with foreign investors and customers.