JIAYI (WESLEY) XU   徐嘉亿

London · Shanghai · Nanjing  |  wesleyxu0622@gmail.com  |  136-0515-9615

AI Product Manager  •  LLM Engineer  •  AI Agent  •  Developer Tools  •  AI Coding

Available for internship: Jun 2026 – Sep 2026 (3+ months)

wesleyxu0622@gmail.com 136-0515-9615 EN Resume 中文简历下载

About Me

UCL Computer Science student with hands-on experience in AI product design, LLM application development, and backend engineering. Focused on AI Agent, AI Coding, Developer Tools, and LLM infrastructure.

Previously held dual PM + LLM Engineer roles at EuniAI and AIMOVerse (now BitRouter), contributing end-to-end from requirements and product flows to Agent workflow design, API design, and backend deployment. Strong ability to bridge deep technical understanding of LLMs, RAG, Agents, multi-model routing, and OpenAI-compatible APIs with product execution.

Education

UNIVERSITY COLLEGE LONDON

09/2023 – 06/2027

MEng Computer Science

Skills

Product

PRD Writing User Research User Flow Design Competitive Analysis Feature Prioritization AI Product Design Product Documentation

AI / LLM

LLM Applications RAG AI Agent Multi-Agent Systems Prompt Engineering AI Coding Model Routing OpenAI-compatible API Model Evaluation

Technical

Python Java JavaScript / TypeScript C / C++ SQL FastAPI SpringBoot Next.js PostgreSQL MySQL Docker Git Linux

Tools

Claude Code Cursor VSCode IntelliJ IDEA GitHub Postman Notion / Feishu Docker

Experience

Product Manager · LLM Engineer · Researcher

EuniAI

Jun 2025 – Oct 2025

  • Contributed to the design and development of Prometheus, an auto-debugging Coding Agent product supporting code understanding, issue classification, bug reproduction, patch generation, and automated verification.
  • Defined Auto Debug user flows and product requirements; decomposed tasks into context retrieval, error localization, patch generation, test validation, and user feedback loops.
  • Participated in multi-agent workflow, API, test validation environment, and Docker-isolated execution environment design, improving Agent output controllability and reliability.
  • Co-authored the Prometheus product paper, covering Auto-Debug workflow, multi-agent collaboration mechanisms, experiment design, and application value.
EuniAI Website Prometheus Repo

Product Manager · LLM Engineer

AIMOVerse (now BitRouter)

Oct 2024 – Jun 2025

  • Participated in the design and development of an LLM Router / AI Marketplace platform supporting unified multi-model invocation via OpenAI-compatible API.
  • Mapped core user journeys for API users, model providers, and AI Agents — covering onboarding, authentication, model invocation, billing, and usage tracking.
  • Contributed to Provider Routing, Proxy/Router configuration, provider onboarding, and deployment debugging; translated availability, latency, and pricing factors into routing strategies.
  • Converted on-chain settlement, pay-as-you-go, and model routing mechanisms into product features and developer documentation to lower integration barriers.
View BitRouter

Backend Developer · Product Feature Design

Jamboxx-Infinite · UCL IXN Voice AI Project

Oct 2024 – Mar 2025

  • Built a Voice Cloning / Voice Conversion RESTful API integrating PyTorch, TorchAudio, DDSP-SVC, and Diffusion models.
  • Designed voice conversion feature flows and model interface encapsulation, supporting frontend invocation, result return, and downstream productization.
Project Website Source Code

Projects

MatrixArena: Dynamic Multi-Dimensional LLM Coding Arena

Mar 2026 – Present

Role: Project Lead / Researcher / Core Developer.

Research project in collaboration with UCL SOLAR to build a dynamic peer-evaluation arena for LLM coding ability and reduce benchmark contamination, overfitting, and single-metric evaluation bias.

Designed and implemented a 5-stage pipeline: Generator → Solver → Sandbox Execution → Judge → Elo Update, enabling model-generated tasks, multi-model parallel solving, cross-rating, and dynamic Elo ranking.

Currently improving evaluation stability, fairness mechanisms, sandbox execution reliability, and scoring consistency while contributing to related academic writing.

GitHub Repo

Prometheus: Auto-Debug Coding Agent

Jun 2025 – Oct 2025

Organization: EuniAI. Role: Product Manager / LLM Engineer / Research Contributor.

Participated in product design and development of an Auto-Debug agent pipeline across code understanding, issue classification, context retrieval, bug reproduction, patch generation, automated test validation, and feedback loop closure.

Contributed to multi-agent workflow design, API and test validation environment, Docker-isolated execution design, and academic writing around auto-debug architecture and experiment methodology.

Website GitHub Repo

BitRouter: Decentralized LLM Routing / AI Marketplace

Jan 2025 – Jun 2025

Organization: AIMOVerse (now BitRouter), an AI startup building infrastructure for AI Agents and developers. Role: Product Manager / LLM Application Engineer.

Participated in platform product design and engineering for OpenAI-compatible unified model invocation, with Provider Routing to optimize availability, latency, cost, and stability across providers.

Owned core user-path design for API users, model providers, and AI Agents (onboarding, auth, model invocation, billing, usage tracing); also worked on Proxy/Router configuration, provider onboarding, deployment debugging, compatibility troubleshooting, and developer documentation.

Translated on-chain settlement, pay-as-you-go billing, and routing mechanisms into product features that lowered developer integration barriers.

Website GitHub Repo

AIMO-Models: Emotional Companion LLM Backend

Oct 2024 – Mar 2025

Role: Product Design / Backend Development / Model Training.

Built and iterated on backend capabilities for emotional companion scenarios with a lightweight API service supporting deployment on CPU/GPU environments and Dockerized runtime.

Participated in model training and backend integration, plus LiteLLM Proxy based multi-provider routing with health checks, model listing, and chat completion proxy endpoints.

Contributed to environment setup, deployment workflows, API usability, and test/coverage practices to improve system reliability and developer onboarding.

GitHub Repo

Jamboxx-Infinite – UCL IXN Project

Sep 2024 – Present

Building RESTful API service for Voice Cloning and Voice Conversion with FastAPI.

Deep Learning: PyTorch, TorchAudio-based DDSP-SVC and Diffusion models for voice conversion.

High-Performance Computing: CUDA GPU acceleration support.

Deployment Engineering: Nuitka compilation to standalone executable.

Project Website Source Code

UCL Engineering Challenge Group Project: Bioreactor Control System

Jun 2023 – Aug 2023

Designed and implemented the string subsystem using Arduino and ESP32, enabling integration between sensors, actuators, and the control system.

Solved communication issues between ESP32 microcontroller and the ThingsBoard cloud platform.

Optimized embedded system performance with MQTT-based IoT communication and C++ device-to-cloud engineering.

Source Code