PENGYI PENGResearch + Engineering
Professional portrait of Pengyi Peng

PENGYI / 2026

Pengyi Peng

AI-native systems builder, quantitative researcher, and open-source engineer.

MSc Complex Systems Modelling, Mathematics
King's College London · Shenzhen, China

EmailGitHubLinkedInCV
AI Agent SystemsQuantitative ResearchFull-Stack EngineeringOpen-Source Contribution

01 / ABOUT

Building systems that make ideas testable.

I work across artificial intelligence, quantitative research, and software engineering. My background in mathematics and complex systems shapes how I model state, uncertainty, feedback, and failure before turning an idea into software.

My current work focuses on AI-native engineering workflows, research infrastructure, trading and risk systems, and public artifacts that expose real code, tests, deployment, and review evidence.

Open to research collaboration, PhD conversations, AI engineering, quantitative research, and long-horizon product building.

02 / RESEARCH

Current research agenda

Three connected directions, organized around reproducible systems rather than isolated demos.

R01

Agent systems

Tool use, memory, retrieval, planning harnesses, evaluation, Human Gates, and multi-model orchestration.

  • Capability routing
  • Evaluation protocols
  • Governed execution
R02

Quant systems

Research pipelines that connect data, signals, portfolio construction, risk, execution, and attribution.

  • Fixed income and factors
  • Market microstructure
  • Research-to-production
R03

Software infrastructure

Typed interfaces, state models, observability, tests, deployment, and architecture-preserving translation across languages.

  • Full-stack systems
  • Polyglot design
  • Evidence-driven delivery

03 / SELECTED PROJECTS

Systems in public

Selected working artifacts with a live surface, source, or both. The wider private research registry is intentionally excluded.

FULL-STACK / LIVE2026

PWEB

A React, TypeScript, Hono, Cloudflare Workers, and D1 system for learning, building, and managing web products through evidence gates.

ReactTypeScriptWorkersD1
Live systemSource
CAREER INFRA / LIVE2026

PENGYICV

An AI-native CV system built around canonical TeX sources, evidence-bounded variants, reproducible packages, and deployable presentation surfaces.

TeXEvidenceAutomation
Open project
QUANT / LIVE2026

Rates Bond Quant

An interactive fixed-income laboratory for duration, convexity, DV01, curve behavior, and rates research.

FICCRiskInteractive
Open laboratory
AGENT PLUGIN / SOURCE2026

dsh-quant

A DeepSeek Harness plugin surface for bounded, agent-native quantitative research workflows.

AgentsQuantPlugin
Source

04 / OPEN SOURCE

Selected merged contributions

Concrete upstream outcomes, linked directly to maintainer-reviewed pull requests.

HKUDS / LightRAGSurface token-limit truncation and fail on empty truncated responsesMerged · #3607 NautilusTraderRemove redundant fixed waits from Coinbase request testsMerged · #4637 Tencent / WeKnoraExclude archived Wiki pages from active statistics queriesMerged · #2549

05 / PATH

Experience and education

Ping An Bank, Shenzhen Branch

Management Trainee · Corporate finance operations and industry research exposure.

King's College London

MSc Complex Systems Modelling · Department of Mathematics.

University of Reading / NUIST

BSc Mathematics and Applied Mathematics.

06 / CONTACT

Research, build, and create durable evidence.

For research collaboration, PhD conversations, AI and quantitative roles, product collaboration, or funding discussions:

pengpengyi92@gmail.com
GitHubLinkedInCV