Resume of Jinlong Peng
Personal Information
| Name: Jinlong Peng | Sex: Male | Age: 30 |
| Phone: 15025586467 | Email: long2ice@gmail.com | Position: Full-stack Engineer |
Professional Skills
- Programming Languages: Proficient in Python, Go, and JavaScript/TypeScript.
- Databases: Experienced with MySQL, Redis, PostgreSQL and other relational databases, used ClickHouse, StarRocks and other OLAP databases.
- Containerization: Skilled with Docker, docker-compose, and Kubernetes, with production deployment experience.
- CI/CD: Experienced with GitHub Actions, GitLab CI/CD and other continuous integration and delivery tools.
- Web Frameworks: Proficient in web frameworks like FastAPI, Django, Gin, and Fiber, as well as front-end frameworks like React and Next.js.
- AI: Proficient with Codex, Claude Code and other AI coding assistants; experienced building agent applications with large language models.
Education
- Chongqing University 2014-9 ~ 2018-7 Computer Science - Undergraduate
Work Experience
Virtue AI 2025-12 ~ 2026-07 Agent Development Engineer (Remote)
- Designed and built AI-security agents for enterprise agent red-teaming and runtime protection.
- Contributed to the MCP Gateway for unified MCP tool access and governance, including tool scanning, access control, and traffic observability.
- Worked on AgentSuite / VirtueGuard capabilities covering attack simulation, runtime interception, and policy-driven safeguards.
- Built automated security agents on large language models for red-team orchestration, risk detection, and result analysis.
Beijing Merico Technology Co., Ltd. 2021-11 ~ 2025-11 Backend Engineer (Remote)
- Led the system design and implementation of the code analysis engine using Python, FastAPI, and Celery.
- Built backend services for Merico's enterprise engineering analytics platform with Node.js, NestJS, and PostgreSQL.
- Resolved production issues for enterprise customers and delivered tailored technical solutions.
- Independently developed an internal test environment management platform to automate environment provisioning and maintenance, improving development and testing efficiency.
- Built a code completion plugin and an automated testing system using AI large models.
Chongqing Mekelaifu Yire Technology Co., Ltd. 2021-03 ~ 2021-11 Backend Engineer
- Delivered core backend features and owned the unified messaging platform.
- Launched a short-link service with Gin and MongoDB plus a distributed ID generator.
- Built services with FastAPI and TortoiseORM in Python, and Gin with GORM in Go.
- Operated infrastructure based on MySQL, Redis, and message queues.
Chongqing Guangleng Technology Co., Ltd. 2019-04 ~ 2021-02 Backend Team Lead
- Oversaw backend architecture, operations, monitoring, and technical roadmap.
- Introduced Kubernetes on Rancher for load balancing and cluster management.
- Adopted ClickHouse for analytics and open-sourced Synch to integrate real-time data with Metabase.
- Established Jenkins-based CI pipelines with automated testing and container image publishing.
- Rolled out a self-hosted Sentry stack for real-time error tracking with instant email notifications.
- Rebuilt services with FastAPI and TortoiseORM, created the Aerich migration tool, and became a TortoiseORM core maintainer.
- Authored and open-sourced FastAPI-Admin as a modern alternative to Django Admin.
Chongqing Xianggua Technology Co., Ltd. 2018-07 ~ 2019-04 Backend Engineer
- Built and operated backend services, including API design and documentation.
- Delivered projects on Django and Flask with MySQL and Redis.
- Containerized workloads with Docker and automated delivery through Jenkins.
Side Projects
Teleguard
An LLM-powered ad-blocking bot for Telegram groups, written in Rust with teloxide and running in production (@TeleguardAdBot).
- Three-tier verdict pipeline — exact Redis cache → pgvector semantic cache (BGE-M3) → LLM — with single-flight collapsing concurrent requests, sharply cutting LLM calls and latency.
- Ordered provider chain with failover across Anthropic, OpenAI, and local compatible endpoints, tracked by latency and success rate.
- False-positive feedback loop: one admin tap restores the user and feeds the sample back as a few-shot example, so verdicts keep self-correcting.
- Admin console embedded into the single binary at compile time via rust-embed — no extra deployment artifacts.
SmsLLM
An iOS scam-SMS filter shipped on the App Store, built with a Rust + Axum backend, SwiftUI, and an ILMessageFilterExtension.
- Designed an offline-classification architecture with out-of-band rule updates around Apple's sandbox (the extension cannot reach the network), so editing rules takes effect immediately without an App Store release.
- A signal-weighted scoring engine replaces keyword matching, with an OTP/2FA fast-path guard that sharply reduces false positives.
- The engine is language-agnostic behind a language-pack abstraction, shipping Chinese and English out of the box; a new language is one more pack.
- Moved the LLM off the delivery path into offline rule training: reported samples → candidate rules → human review → client re-sync.
MedVault
A family medicine-cabinet manager on the App Store, with a Rust/Axum backend, a SwiftUI iOS app, and a Taro WeChat mini-program.
- The AI assistant runs a tool-calling loop where writes never execute server-side: they are emitted as SSE events for the client to confirm and send through the normal endpoints, so there is no privileged AI write path.
- Chat streams over SSE, routing image turns to a vision model and tool turns to a text model, with an iteration cap guarding against runaway loops.
- Authorization is cabinet-scoped rather than user-scoped, returning 404 instead of 403 for non-members to avoid leaking resource existence.
- One backend API serves both iOS and the WeChat mini-program; StoreKit subscriptions integrate the App Store Server API to self-heal lost webhooks.
Open Source Projects
TortoiseORM 
One of the most popular asyncio ORMs in Python, with a Django-like API. I contributed features such as F-expressions, signals, and GROUP BY support, and now serve as a maintainer.
FastAPI-Admin 
An out-of-the-box admin dashboard for the FastAPI + TortoiseORM stack, inspired by Django Admin and communicating over REST. The project now lives in its own GitHub organization.
fastapi-cache 
A decorator-based cache for FastAPI endpoints and function results, supporting Redis, Memcached, DynamoDB, and in-memory backends plus HTTP cache semantics such as ETag and Cache-Control.
Aerich 
A database migration tool I built for TortoiseORM, comparable to Alembic for SQLAlchemy. Developed independently and now the official migration solution under the Tortoise organization.
asyncmy 
A high-performance asyncio MySQL/MariaDB driver with a Cython-powered protocol core, an aiomysql-compatible API, a built-in connection pool, and binlog replication support.
meilisync 
Real-time sync from MySQL, PostgreSQL, and MongoDB to Meilisearch with both full and incremental modes, paired with the meilisync-admin dashboard.
Other Links
- GitHub: https://github.com/long2ice
- Blog: https://blog.long2ice.io
- Chinese Resume: https://resume.long2ice.io
- English Resume: https://resume-en.long2ice.io