About

Zesheng Cai (Jaycen)

I’m Zesheng Cai (Jaycen), a Tokyo-based software engineer focused on backend systems and AI applications. My experience spans marketplace services at Rakuten, enterprise document intelligence and data platforms at KPMG FAS, and advertising reporting at SmartNews.

I build APIs, data pipelines, and cloud-native applications with Python and JavaScript/TypeScript. I also develop personal projects exploring AI agents, retrieval, and persistent memory for coding assistants.

I hold a master’s degree in Information Engineering from Osaka University.

Based in
Tokyo, Japan
Languages
Chinese (native), Japanese (business, JLPT N1), English (business)
Email
saitakusei@gmail.com

Education

Master of Information Engineering

Apr. 2019 - Mar. 2021
Osaka University
  • Machine Learning

Bachelor of Computer Science and Engineering

Sep. 2014 - Jun. 2018
Sun Yat-sen University
  • Computer Vision

Certifications

CKAD

Cloud Digital Leader

Experience

SmartNews

Software Engineer · Ads Product / Ads Reporting | October 2026 – Present

  • Work on the Ads Reporting team supporting the advertising platform.
  • Currently developing an AI-powered campaign performance summary feature that combines reporting metrics with LLM-generated explanations.

KPMG FAS

AI & Data Platforms Engineer | July 2025 – September 2026

  • Promoted to Senior Associate in September 2026.
  • Built an internal AI Virtual Data Room application for document ingestion, summarization, question answering, and semantic search using Python, LangChain, and Azure Document Intelligence.
  • Developed financial-disclosure data processing and backend access for the EDINET platform using Databricks, PySpark, SQL Server, and ORM-backed APIs.
  • Worked with internal stakeholders on technical investigation, prototyping, and hands-on implementation of AI and data tools.

Rakuten Group

Software Engineer · Car Auction | April 2021 – June 2025

  • Developed marketplace backend services and APIs using Node.js, Python, GraphQL, and MySQL, alongside Vue.js and Nuxt.js frontend applications.
  • Designed and implemented email-processing and Outlook-scheduling microservices using Microsoft Graph API to replace Google Apps Script workflows.
  • Integrated car-grade data using Python, Elasticsearch fuzzy search, and MySQL to reconcile inconsistent vehicle naming.
  • Maintained GCP infrastructure with Kubernetes and Terraform, and automated testing and deployment workflows with GitHub Actions.

Personal projects

Selected personal projects in agent tooling and AI applications, developed independently of my employment.

Agentic Memory MCP

Coding-agent infrastructure · Python / FastMCP / SQLite / Chroma

  • Persistent task memory for coding agents, combining BM25 and vector retrieval with reranking and progressive context disclosure.
  • Background trace distillation, memory merging, feedback, and recoverable archival.
  • Local MCP server; integration with my separate agent harness is planned.

Project Management Agent

Workflow automation · OpenAI Agents / MCP / GitHub

  • Turns natural-language requirements into GitHub issues and project-board actions through a custom MCP server.
  • Separates requirement analysis from execution with structured handoffs and human approval steps.

Multi-Agent PPT Generation

AI application · LangChain / FastAPI / NestJS / React

  • Chat-driven workflow for planning, generating, and refining presentations with a custom Marp theme.
  • Separate agent and rendering services generate presentation files, with a React interface for iterative refinement.

Explore more of my work on my project portfolio or GitHub.

Skills

Backend & web

Python, JavaScript/TypeScript, Node.js, Express, FastAPI, GraphQL, REST APIs, SQLAlchemy, Vue.js, Nuxt.js

AI applications

RAG, LangChain, document extraction, Azure Document Intelligence, semantic search, LLM summarization; MCP and agent workflows through personal projects

Data & search

MySQL, SQL Server, PostgreSQL, Redis, Elasticsearch, Databricks, PySpark

Cloud & delivery

GCP, GKE, Cloud Run, Azure Blob Storage, Docker, Kubernetes, Terraform, GitHub Actions, Jest

Currently developing: deeper expertise in Kafka, Flink, ClickHouse, Airflow, and distributed streaming systems through Ads Reporting work; exploring agent orchestration and LLM evaluation.

Contact

Location

Tokyo, Japan

Social Profiles

Email

saitakusei@gmail.com

Resume

Request my resume