Skip to content
J.R.

J.R.

Full-Stack Engineer / AI & Backend

Building AI products since 2019, centered on machine learning and production-grade backend engineering. Currently in Japan working on AI code-analysis and business systems.

View projects ↓ Skill sheet (Excel)
Scroll

ABOUT

About

My strength is a strong curiosity toward new things and a willingness to take on challenges. Whenever a new technology or tool appears, I learn it and try it out hands-on first. As a result I pick things up quickly and adapt flexibly to new environments. I welcome change and keep growing while enjoying new opportunities.

Education

  • Hebei University of Economics and Business (B.Sc.)

    Mathematics and Applied Mathematics

    2015.09 – 2019.06 ・ GPA 3.97 / 5 · Rank 9 / 30 in major

Profile

Gender
Male
Age
29
Nationality
China
Nearest station
Kameari Station
Arrived in Japan
2024.07
Experience
6 yrs 9 mos
Location
Tokyo, Japan
Email
cifar666gmail.com

SKILLS

Technical Skills

Python Backend

Building and maintaining web applications with Django, Tornado and Flask. Data processing and storage with MySQL, MongoDB, Redis, Kafka and Milvus. Interface design and integration over RESTful APIs and Protobuf-based gRPC.

Machine Learning & Computer Vision

Data analysis (regression, clustering) with scikit-learn. Object detection (Faster R-CNN), recognition (ResNet) and YOLOv8 with PyTorch. Inference acceleration on NVIDIA CUDA and Huawei Ascend ACL with C++ integration.

Frontend

Admin UIs with Vue 2 / Vue-Element-Admin. Screen development with React, TypeScript and Angular. Data visualization with ECharts.

Multiple Languages

C++11 project work. API and business-logic implementation in Go. Experience with a migration project to Rust.

Deployment & Ops

Provisioning and deployment with Docker, docker-compose, Nginx and shell scripts. Production setup and operations on AWS and Azure. Handling vulnerability scans with tools such as Fortify.

Data Processing & Misc.

Crawler development (XPath-based and API-based). Data manipulation with NumPy and Pandas. Image processing with OpenCV; video rotation, affine transforms and cropping with FFmpeg. Applied LLM / prompt engineering at work, including building an internal GPT bot.

Experience Matrix

◎: 1+ year professional / ○: professional experience / △: familiar

OS

  • Windows ◎
  • Windows Server ◎
  • Linux ◎
  • macOS ◎
  • AWS ◎
  • Azure ◎

Language

  • Python ◎
  • Vue.js ◎
  • React.js ◎
  • Angular.js ◎
  • TypeScript ◎
  • JavaScript ◎
  • HTML ◎
  • jQuery ◎
  • AJAX ◎
  • Go ◎
  • Shell Script ◎
  • C ○
  • C++ ○
  • VC++ ○
  • COBOL △

Database

  • MySQL ◎
  • PostgreSQL ◎
  • SQL Server ◎
  • SQLite ◎
  • MongoDB ◎
  • Redis ○
  • Milvus ○

Web / AP Server

  • Nginx ◎
  • Apache ◎
  • IIS ◎
  • Tomcat ○

Other

  • Docker / docker-compose ◎
  • Kafka ◎
  • PyTorch ◎
  • scikit-learn ◎
  • OpenCV / FFmpeg ◎
  • gRPC / Protobuf ○
  • Ray ○
  • Selenium ○
  • LLM / Prompt Engineering ○
  • CUDA / Huawei Ascend ACL △

Summary

Focus areas
Deep learning (CV)Machine learningData analysisWeb developmentWeb crawling
Languages
Python (professional)Vue / React / TypeScript (professional)GoC++Rust (familiar)
Infrastructure & middleware
MySQL / MongoDB / PostgreSQL / RedisKafka / MilvusDocker / docker-compose / NginxAWS / Azure

Languages

  • Chinese Native

    Mother tongue

  • Japanese JLPT N2 / basic business conversation

    Self-taught, ~2 years

  • English Technical reading & writing

Certifications

  • 2026.07 JLPT N2 (Japanese-Language Proficiency Test)

EXPERIENCE

Work Experience

  1. 2024.07 – Present

    ◯◯ Co., Ltd. (SES)

    Full-Stack Engineer ・ Contract development (Japan)

  2. 2022.08 – 2023.07

    Zhejiang Mulian Technology Co., Ltd.

    Python Developer ・ Industrial control security

  3. 2020.08 – 2022.08

    People's Zhongke Intelligent Technology Co., Ltd.

    Algorithm Application Engineer ・ AI × content safety (public-opinion monitoring)

  4. 2019.07 – 2020.07

    Beijing Yunshangzuo Technology Co., Ltd.

    Software Engineer ・ AI × education

How I Got Here

  1. 01

    During university (mathematics): learned data analysis with Python/scikit-learn and NumPy/Pandas. Amid the face-recognition boom, studied the C++ OSS face-recognition framework SeetaFace.

  2. 02

    After joining AI × education: implemented Faster R-CNN and ResNet with PyTorch inside business logic. Web moved to Tornado; the frontend moved from HTML/JS/CSS to Vue, broadening my CS foundation.

  3. 03

    AI × public-opinion period: worked with CV, OCR and NLP; used Kafka to unify multiple platforms in the core algorithm program, and ran Docker and cloud operations. Learned model distillation/pruning and hardware-level acceleration on CUDA / Ascend.

  4. 04

    Security industry: understood common attack techniques and privilege escalation. Became proficient with Django; took part in a C++-to-Rust migration and Python speedups via Cython or rewrites in Go.

  5. 05

    In Japan: working on AI code analysis and business systems, doing full-stack development aligned with Japanese development processes.

WORKS

Project History

Japan 2026.01 – 2026.03 （3 mos）

Internal Project-Management System (Redmine customization)

Manufacturing (major electronics maker) ・ Full-Stack Engineer (PG) ・ Team 5

Details
  • Python
  • Angular
  • Go
  • Redmine
  • Selenium
  • Fortify
  • +3

Development of an internal project-management system for a major electronics manufacturer. Based on the OSS Redmine, we built custom cycle-management and progress-visualization features to fit the client's requirements, including role-based view permissions and department-level progress tracking.

Responsibilities

  • Screen changes and new features in Angular
  • API and business logic in Go
  • Redmine feature extensions and permission-control customization
  • Data integration with the internal OA system
  • Employee-data fetch and automated register/update batch jobs in Python (Selenium)
  • Code review, vulnerability-scan remediation, quality improvement via static analysis

Outcomes

  • ✓ Automated employee-data registration, cutting operational effort
  • ✓ Enabled department-level progress visibility through role-based access control
  • ✓ Delivered a management platform matching internal requirements in about 3 months by leveraging OSS
Scope: 機能設計 / 内部設計 / 製造 / 単体試験 / 結合試験
  • Python
  • Angular
  • Go
  • Redmine
  • Selenium
  • Fortify
  • Azure
  • Docker
  • Git
Japan 2025.04 – 2025.12 （9 mos）

AI Code-Analysis Service (visualizing legacy COBOL assets)

AI (legacy-asset visualization & documentation) ・ Systems Engineer ・ Team 3

Details
  • Python
  • React
  • COBOL
  • LLM
  • Prompt
  • Nginx
  • +3

An AI-assisted system that analyzes existing COBOL assets, visualizes their structure and auto-generates documentation. We implemented LLM-based code summarization and design-document generation to speed up investigation and maintenance work.

Responsibilities

  • Design and development of COBOL source analysis (format conversion, structure parsing, etc.)
  • Design-document templates and automated documentation generation
  • Design and build of web/desktop screens to visualize analysis results
  • API interface design and implementation with external AI services
  • Team development process setup and standardization

Outcomes

  • ✓ Cut investigation effort by 30–50% by making structure and impact analysis easier
  • ✓ Greatly reduced design-doc writing/review effort and standardized quality
  • ✓ Established a source-analysis + AI pipeline extensible to other languages
Scope: 外部設計 / 機能設計 / 内部設計 / 製造 / 単体試験 / 結合試験
  • Python
  • React
  • COBOL
  • LLM
  • Prompt
  • Nginx
  • AWS
  • Docker
  • Git
Japan 2024.07 – 2025.03 （9 mos）

Property-Management System for Vacation Rentals

Real estate (vacation rental) ・ Systems Engineer ・ Team 3

Details
  • Python
  • Flask
  • React
  • TypeScript
  • PostgreSQL
  • Nginx
  • +3

A back-office system for vacation-rental operators, offering public property listings and an admin-facing property management console. Booking and payment are handed off to Airbnb's external site. I handled everything from server setup to production deployment and operations.

Responsibilities

  • Property management features (create/edit/delete/list): development, unit tests, maintenance
  • Public-facing property listing display
  • Airbnb integration for booking and payment
  • Inquiry handling and minor spec changes
  • Server provisioning, production deployment and operations
Scope: 外部設計 / 機能設計 / 内部設計 / 製造 / 単体試験 / 結合試験
  • Python
  • Flask
  • React
  • TypeScript
  • PostgreSQL
  • Nginx
  • AWS
  • Docker
  • Git
China 2022.11 – 2024.06 （1 yr 8 mos）

Classroom Behavior Analysis System

AI × education ・ Full-Stack Engineer (SE) ・ Team 5

Details
  • Python
  • Tornado
  • scikit-learn
  • Bootstrap
  • Ajax
  • jQuery
  • +3

An education solution that uses AI to gauge students' attention in class, digitizing lesson evaluation and giving feedback to parents. Student actions (leaning forward, smiling, turning away, standing, slumping, raising a hand, etc.) are classified automatically, streamlining teacher evaluation and letting parents review their child's classroom behavior afterward.

Responsibilities

  • Backend (Python / Tornado): designed and built API features by composing existing AI model functions from the prototype
  • Defined and implemented API interface specs (protocol and field design)
  • Behavior-data analysis with scikit-learn (applying the attention-classification model)
  • Per-student report generation (weekly, monthly, term reports)
  • Frontend (Bootstrap): revised and extended the class-data visualization UI in the admin console
  • Report display in the parent-facing mini-program (WeChat integration)
Scope: 外部設計 / 機能設計 / 内部設計 / 製造 / 単体試験 / 結合試験
  • Python
  • Tornado
  • scikit-learn
  • Bootstrap
  • Ajax
  • jQuery
  • MongoDB
  • Postman
  • Git
China 2021.10 – 2022.10 （1 yr 1 mo）

Content-Safety Review System

AI × content safety (public-opinion risk) ・ Platform / Algorithm Backend Engineer (SE) ・ Team 15

Details
  • Python
  • PyTorch
  • Ray
  • ResNet
  • YOLOv8
  • Kafka
  • +5

Aimed at an enterprise-grade content-safety review platform: it monitors and analyzes media content on the public internet and provides safety-review capabilities. Models (face detection, logo, scene, OCR, ASR, cross-modal, etc.) are consumed through a subscription model.

Responsibilities

  • Maintenance and feature extension of existing algorithm-program modules
  • Processed task data produced by upstream microservices via a Kafka queue
  • Consumed Protobuf messages from Kafka and extracted the corresponding video/image/text
  • Generated model results by subscribed algorithm ID and returned responses per the interface spec
  • Handled protocol changes, field additions/removals, algorithm upgrades and parameter-format changes
  • Delivered offline products: wrapped existing capabilities into simplified customer-facing interfaces
  • Example — firearm detection from video: extract frames from a video URL → generate 512-dim features via the model SDK → query Milvus over gRPC → post-process and return recognition results → ship as a Docker image

Data flow

Upstream apps → Kafka (Protobuf tasks) → algorithm program → model inference → results stored in Milvus / MongoDB → results returned via API

Scope: 外部設計 / 機能設計 / 内部設計 / 製造 / 単体試験 / 結合試験
  • Python
  • PyTorch
  • Ray
  • ResNet
  • YOLOv8
  • Kafka
  • Milvus
  • PyMongo
  • gRPC
  • Docker
  • Git
China 2020.05 – 2021.11 （1 yr 7 mos）

Industrial-IoT Product Audit System

Industrial IoT × security ・ Full-Stack Engineer (SE) ・ Team 13

Details
  • Python
  • Django
  • Suricata
  • libpcap/sniff
  • MySQL
  • Vue 2
  • +2

An audit system that protects the networks of PLCs and control equipment in industries such as energy and chemicals, in line with China's Cybersecurity Law and its classified-protection scheme. It monitors all factory-network traffic, detecting vulnerabilities and raising security alerts.

Responsibilities

  • Maintenance of business modules: updated business-side logic as the protocol-parsing library gained/changed protocols
  • New feature API interfaces per frontend requirements
  • Split and recomposed the audit system into modules to deliver contract-specific custom builds
  • Added features and reworked screens for the web UI, improved the alert display

Data flow

All mirror-port traffic → captured with sniff → industrial-protocol parsing (CVE/CNVD matching via Suricata) → IP/MAC/port/payload stored in MySQL → policy check (allow/deny lists) → alerts pushed to the web UI

Scope: 外部設計 / 機能設計 / 内部設計 / 製造 / 単体試験 / 結合試験
  • Python
  • Django
  • Suricata
  • libpcap/sniff
  • MySQL
  • Vue 2
  • JavaScript
  • Git
China 2019.07 – 2020.04 （10 mos）

CV Data-Processing System

AI × education (data platform) ・ Systems Engineer ・ Team 1

Details
  • Python
  • Tornado
  • Vue
  • Vue-Element-Admin
  • MongoDB
  • OpenCV
  • +1

Built to visualize and streamline the data-processing work that deep learning needs, letting team members collect and transform data through a GUI. Scattered personal scripts were consolidated into one system, improving reuse and efficiency.

Responsibilities

  • Data-collection module: integrated crawler services over RESTful APIs, ran collection tasks from the UI, stored results in MongoDB
  • Data-processing module: exposed OpenCV image ops and FFmpeg video ops as Tornado APIs, operable from a Vue UI; designed resumable chunked uploads
  • Model-defect analysis module: GUI for initial labeling and human review; computed PR metrics across datasets/models and auto-generated PDF reports
  • Worked across frontend and backend, including page fixes and UI improvements
Scope: 要件定義 / 外部設計 / 機能設計 / 内部設計 / 製造 / 単体試験 / 結合試験
  • Python
  • Tornado
  • Vue
  • Vue-Element-Admin
  • MongoDB
  • OpenCV
  • FFmpeg

Other projects

  • Live-stream moderation appliance
  • Style-image generation (Midjourney-like)
  • Vehicle behavior analysis
  • Congenital-disease screening from facial images
More projects (on Gitee) ↗

ARCHIVE

Early Work — 2019

Things I built on my own from university through my first year working (around 2019) — the period when I moved from a maths major into ML and web development by self-study. The full site is preserved as-is at /2019/.

Classroom-analytics product (admin UI)

Classroom-analytics product (admin UI)

For an English-conversation school: a dashboard visualizing student attention and teacher metrics from lesson videos.

  • Vue
  • ECharts
  • Tornado
Data wall / big screen

Data wall / big screen

A big-screen view summarizing metrics across classrooms, teachers and students.

  • Vue
  • ECharts
Per-student report

Per-student report

Aggregated behavior metrics (attendance, smiling, hand-raising, …) per student, generating weekly/monthly reports.

  • scikit-learn
  • Tornado
In-class behavior classification

In-class behavior classification

Image classification of postures — facing front, looking down, turning away, slumping, standing, etc. — trained and applied with ResNet.

  • PyTorch
  • ResNet
Action recognition from video

Action recognition from video

Used 3D-ResNet for classifying human actions that depend on temporal context.

  • PyTorch
  • 3D-ResNet
Keypoint detection (hand-raise / sit-up count)

Keypoint detection (hand-raise / sit-up count)

Human keypoints via Detectron2, used for hand-raise detection and counting sit-ups.

  • Detectron2
  • PyTorch
Facial analysis system

Facial analysis system

A prototype MVC web system doing facial-landmark detection and feature extraction with Dlib.

  • Dlib
  • Tornado
  • MongoDB

Product demo clips

Walkthrough clips of the classroom-analytics product at the time (3 clips).

  • Vue
  • Tornado

Awards

  • 2018 China Undergraduate Mathematical Contest in Modeling — 2nd prize, Hebei province
  • 2018 MCM/ICM (Mathematical Contest in Modeling) — Successful Participant
  • 2018 National Market Research & Analysis Contest — 1st prize, Hebei province
  • 2017 Inner Mongolia Mathematical Modeling Contest — Participant / awarded
  • 2016–2019 Academic scholarship — 3rd class (sophomore–senior), plus a diligence award

The above is only a selection. Study notes, detailed screenshots and my profile at the time are all preserved on the 2019 site.

2019 · resume-master

Open the 2019 version ↗

CONTACT

Get in Touch

For work inquiries, please reach me by email.