Peter Guan

Transferring Fall 2027. Relevant Coursework: C++ Programming, Pre-Calculus. Active member of the CS Club.

Completed two years of rigorous coursework including Software Engineering, Data Structures & Algorithms, and Reasoning & Agents (AI).

Explored the intersection of Engineering and Finance. Withdrew to pursue specialized CS studies at Edinburgh.
Award: International School Leaver Scholarship (2022).

• Leading a research initiative with UCLA HCI faculty to investigate how eXplainable AI (XAI) impacts user trust and empathy in emotional support LLMs.
• Designed and executed an A/B study, developing a custom evaluation platform to compare baseline LLMs against models integrated with real-time sentiment analysis and XAI transparency.
• Synthesizing quantitative and qualitative data to prepare a peer-reviewed publication as the sole author (Expected Feb 2026).

• Architected "Veru", an AI citation auditor that cross-references LLM outputs against 250M+ academic papers (OpenAlex) to detect hallucinations.
• Engineered a modern full-stack pipeline using Next.js 14 (Frontend) and FastAPI (Backend), integrating Gemini 2.0 Flash for high-speed semantic extraction.
• Implemented a multi-tier verification strategy, combining OpenAlex, Semantic Scholar, and Google Search Grounding to achieve forensic-level accuracy.
• Launched to production on Vercel/Render, serving 100+ unique users in the first week with zero downtime.

• Architected a technical pipeline to model the effects of behavioural biases (FoMO, Herding) on excessive trading.
• Developed a comprehensive data analysis workflow using R for PLS-SEM modelling and Python (pandas) for data engineering, bridged via rpy2 in a WSL environment.
• Validated the hypothesis that Fear of Missing Out (FoMO) is a primary predictor of irrational trading behaviours through statistical significance testing.

• Spearheaded backend development for "Realibuddy," a real-time conversational fact-checking assistant.
• Integrated Deepgram ASR and Perplexity API via WebSocket for low-latency voice processing and grounded truth retrieval.
• Standardized API communication between the Python backend and React frontend, ensuring seamless real-time alerts.
• Won Best AI/ML Award (Top 1 out of 29 teams) for technical complexity and product viability.

Developed a custom markup language IDE using JavaFX and ANTLR4. Implemented the Visitor pattern to compile ASTs into HTML/LaTeX and applied MVVM architecture for robust state management.