Skip to content
View LQ458's full-sized avatar
🏠
Working from home
🏠
Working from home

Organizations

@Oasisteens

Block or report LQ458

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
LQ458/README.md

Leo Qin

Incoming Computer Engineering student at Washington University in St. Louis. I build AI systems, agent safety evals, RAG pipelines, and reliability infrastructure for tool-using AI.

What I'm Working On

Daily Admin Agent Security Eval — AgentDojo-compatible eval suite for tool-using daily-admin agents. 5 user tasks, 8 injection tasks, synthetic tools, preflight checks, and OpenAI-compatible model runners. Tests whether agents can complete useful tasks while treating retrieved app content as untrusted data.

LawAI — Privacy-aware RAG legal assistant. Enforces document-level access via Auth0 FGA (HR sees salary docs, engineers denied). RAG pipeline with Pinecone + multilingual-e5-large embeddings. Includes AI evaluation framework using DeepSeek judge to score responses on safety, specificity, and actionability. Next.js 15 + DeepSeek V4 Pro.

Lesson Plan Generator — RAG system for 95K+ curriculum chunks. ChromaDB + Qwen LLM. End-to-end pipeline with OCR error correction and semantic search. Deployed in rural Guizhou schools.

StatsifyFinance — AI financial analysis. DeepSeek V3 streaming, LangChain orchestration, dual OCR (Tencent + Baidu) for document understanding.

BiphFlora — Plant encyclopedia with image recognition. Built a searchable database of plant species with identification features.

Tech Stack

Languages:    Python, C/C++, TypeScript, JavaScript
AI Safety:    Agent Evals, Prompt Injection Testing, AI Control, Tool-Use Safety
AI/ML:        PyTorch, RAG, Vector Search, LLM APIs, CUDA
Databases:    MongoDB, Redis, Pinecone, ChromaDB
Frameworks:   React, Next.js, Node.js, Express
Tools:        Docker, Linux, Git, Auth0 FGA

Recent

  • Building daily-admin-agent-security-eval, a small AgentDojo-compatible agent safety eval suite
  • BlueDot Impact AI Safety Fundamentals international cohort
  • MLH GenAI Auth0 Privacy-Aware RAG Bot challenge completed
  • 2nd place at Moonshot 48 Hackathon (built a smart white cane with YOLOv5 + GPS + AI voice in 48 hours)
  • AIME qualifier (2022) · 11 AP 5s

Contact

GitHub Streak

Pinned Loading

  1. daily-admin-agent-security-eval daily-admin-agent-security-eval Public

    AgentDojo suite for daily-admin agent security evaluation with simulated dynamic tool workflows.

    Python

  2. LawAI LawAI Public

    An AI-powered legal assistant application designed to provide users with intelligent legal case recommendations, case summaries, and legal consultation services.

    TypeScript 1 1

  3. lesson-plan-generator lesson-plan-generator Public

    A full-stack AI-powered lesson plan generation application built with Next.js and Node.js, featuring an integrated RAG (Retrieval-Augmented Generation) system for more accurate educational content …

    JavaScript 1 1

  4. CourseScheduler CourseScheduler Public

    An intelligent constraint optimization solution that eliminates scheduling conflicts while balancing student preferences, teacher availability, and room capacity.

    Python 1