cv.md
NORMAL
UTF-8
>ls ~/downloads/
>cat /users/peifeng/cv.md | render --html-cv
CURRICULUM_VITAE

Peifeng Wang

Tokyo, Japan[email protected]
>looking forMeaningful work · good team · senior scope · long-term
// SUMMARY

Full-stack engineer and team lead with 10+ years building products from zero to one — across SaaS, e-commerce, robotics, and supply chain. Led a 10-person engineering team at Geetest while shipping production features. Co-engineered autonomous forklift navigation with CMU researchers at BITO Robotics. Architected a warehouse design automation tool at Shucheng AI, collaborating with co-founders from Stanford and University of Michigan. For the past three years, based in Tokyo and working across Japanese startups building SaaS and e-commerce platforms.

Strong in Python, Go, TypeScript, and Elixir. Comfortable owning the full stack — architecture decisions, backend systems, frontend delivery, and DevOps — in small teams where all of it sits at the same table.

On the side, building sky-finance: a local-first RAG pipeline with sentiment-bucketed retrieval, multi-provider LLM routing, and LLM-as-judge evaluation — running at under $0.50/month.

// WORK EXPERIENCES
Full-Stack Engineer
Kichijōji, Tokyo, Japan
Aug 2025 – Present

Shopify-focused e-commerce company. Joined as an early core developer on a new internal SaaS initiative: a timecard management platform developed from its early foundation and currently running in internal pilot. Worked across product requirements, system architecture, real-time workflows, and end-to-end feature delivery.

  • Took ownership of feature design and technical architecture for the timecard SaaS, translating operational requirements into production-ready full-stack implementations.
  • Built a real-time kiosk view with Phoenix LiveView and WebSockets, enabling instant punch-in/out status updates across locations without polling.
  • Developed a live 2D canvas map using sprite graphics and Anime.js to visualize employee state transitions as game-style character movement triggered by punch events.
  • Collaborated closely with the team lead on foundational system design while independently delivering core product functionality from concept to pilot.
Lead Full-Stack Engineer
Shinjuku, Tokyo, Japan
Aug 2024 – Jul 2025

Japanese startup building portfolio and commerce products for creators. Stepped into the Lead Engineer role to rebuild the engineering function and deliver the company's commerce platform end-to-end.

  • Took ownership of architecture, delivery priorities, and team operations as Lead Engineer, rebuilding the engineering team and processes from the ground up.
  • Architected and launched a full commerce platform (Rails / Next.js / TypeScript / Stripe) handling 1,000+ transactions with per-order values ranging ¥10K–¥200K, integrated end-to-end from the Foriio main site through to payment settlement.
  • Maintained and refactored Ruby on Rails APIs supporting 2–3 million daily requests, improving stability and resolving recurring performance bottlenecks.
  • Modernized CI/CD pipelines across 10+ projects and 40+ servers — introduced modular GitHub Actions architecture for new projects, replacing monolithic single-file configs on key pipelines to reduce maintenance overhead.
  • Led a team of 4 engineers — conducting interviews, onboarding, task allocation, and priority definition while personally handling the most complex technical deliverables.
Gennmu Technology
Software Engineer
Nihonbashi, Tokyo, Japan
Jan 2023 – Jul 2024

Tokyo-based staffing and technology firm serving major Japanese enterprises. First role after relocating to Japan — contributed across client-facing Salesforce projects and internal web development.

  • Delivered Salesforce Marketing Cloud low-code implementations for major Japanese clients including AU and UNITED ARROWS, involving SQL, JavaScript, and data analysis workflows.
  • Rebuilt and launched the company homepage using Flask, replacing the existing site with a cleaner, maintainable codebase.
  • Developed a gaming equipment e-commerce platform using Django, delivering a complete codebase for internal handoff.
Full-Stack Engineer / Technical Director
Shanghai, China
Jul 2019 – Feb 2022

Supply chain digitization company incubated by SF Express, China's largest logistics enterprise. Joined as a core engineer and took on Technical Director responsibilities — owning technology selection, architecture, and end-to-end team recruitment alongside a founding team from Stanford and the University of Michigan.

  • Re-architected the company's primary Python-based platform from the ground up after joining — redesigning core systems and establishing new technical foundations across front-end and back-end.
  • Architected and shipped a warehouse design automation tool that generates CAD, 2D, and 3D layout models from parametric web inputs, eliminating manual design cycles for logistics PoC clients.
  • Contributed to a Starbucks demand forecasting project (4,000 stores across China) — responsible for data cleaning pipelines and auxiliary web UI for model input and output.
  • Developed a WeChat Mini Program e-commerce platform for multi-channel sales distribution using Django and Vue.js.
  • Led full-cycle recruitment and managed a team of 5 engineers, owning hiring decisions, onboarding, and technical direction.
Senior Full-Stack Engineer
Shanghai, China
Jul 2018 – May 2019

Robotics company specializing in autonomous forklifts and AGV systems, with a core team primarily from Carnegie Mellon University. Rebuilt the front-end and back-end of two core systems from scratch — a central control platform and a per-AGV on-board system — integrating directly with ROS in a live warehouse environment.

  • Rebuilt the AGV central control system (Flask / Redis / Vue.js) from scratch — implemented a custom WebSocket bridge to ROS topics for real-time bidirectional communication between the web platform and autonomous forklift fleet.
  • Developed a road network design tool using D3.js that constructs warehouse maps from QR code scans, allowing operators to define nodes, roads, and AGV task routes entirely through a web interface.
  • Built a per-AGV on-board monitoring system with direct ROS integration, enabling real-time status tracking and task management for a fleet of 5–10 autonomous forklifts.
  • Led a development team of 3 engineers across both systems.
Full-Stack Engineer / Web Team Lead
Wuhan, China
Jun 2014 – May 2018

Cybersecurity startup best known for CAPTCHA and bot-detection — joined as the 13th employee, grew to ~200 employees and nearly 100 engineers during tenure. First engineering role after a career switch from sales — entered the industry as a self-taught developer and progressed to Web Team Lead within one year.

  • Joined as the sole web developer and became Web Team Lead within one year as the company scaled — ultimately leading a team of 10 engineers across front-end and full-stack projects.
  • Architected and delivered the core B2B customer platform — a user management and API key portal that scaled from under 10K to 200K registered users, rebuilt multiple times (Django → Flask + Django + Vue.js) to keep pace with growth.
  • Developed 10+ production web systems across the full product lifecycle — including three full UI rebuilds of the company homepage, an internal business intelligence dashboard for commercial operations, and web tooling for the machine learning team.
  • Rebuilt the company's primary web infrastructure from Django monolith to a modular Flask + Vue.js architecture, establishing the technical foundation the web team operated on through the company's growth phase.
// PERSONAL PROJECTS
sky-finance
LLM-powered financial intelligence platform
  • Built an end-to-end financial intelligence platform in Python — market/news ingestion, LLM-assisted cleaning, embedding, RAG retrieval, and strategy analysis, running at under $0.50/month in cloud API costs.
  • Implemented hybrid RAG retrieval with PostgreSQL + pgvector, combining BM25 full-text search and HNSW vector search via Reciprocal Rank Fusion, with sentiment-bucketed context to ensure minority signals reach the model.
  • Designed multi-provider LLM routing across Ollama, OpenAI, and Anthropic Claude — structured outputs, prompt caching (~90% cost reduction), and per-call token/cost tracking surfaced in the dashboard.
  • Built an LLM-as-a-judge evaluation pipeline comparing sentiment-bucketed vs. flat retrieval across faithfulness, coverage, and actionability dimensions.
// EDUCATION
Wuchang University of Technology
Environmental engineering
Sep 2007 – Jul 2011Wuhan, China
// PATENTS

WO2021184193A1 - Computer-Aided Warehouse Space Planning

// PROFICIENCY
Python95%
JavaScript / TypeScript92%
Go80%
Elixir75%
Ruby70%
// TECH STACK
FRAMEWORKS
DjangoFastAPIPhoenixNext.jsReactFlaskVue.jsRailsExpo
AI & LLM
Claude APIOpenAI APIOllamaRAGpgvectorLLM Pipelines
DATABASE
PostgreSQLRedisMySQLMongoDB
INFRASTRUCTURE
DockerLinuxGitHub ActionsAWSKubernetesAnsibleGCPAzure
// LANGUAGES
MandarinNATIVE
EnglishBUSINESS
JapaneseBASIC
root@wang's server:~#