JUDY459.TOP

FORWARD DEPLOYED ENGINEER

press any key to skip

JUDY459.TOPONLINE
AGENTIC AI SOLUTIONS — DISCOVERY TO PRODUCTIONMULTI-AGENT ORCHESTRATION OVER MATRIXMS CS @ UNIVERSITY OF CHICAGOOPEN TO FORWARD DEPLOYED / SOLUTIONS ENGINEERING ROLESSHIPPED INTO A CUSTOMER'S PRODUCTION ENVIRONMENTAGENTIC AI SOLUTIONS — DISCOVERY TO PRODUCTIONMULTI-AGENT ORCHESTRATION OVER MATRIXMS CS @ UNIVERSITY OF CHICAGOOPEN TO FORWARD DEPLOYED / SOLUTIONS ENGINEERING ROLESSHIPPED INTO A CUSTOMER'S PRODUCTION ENVIRONMENT

C:\PORTFOLIO\> LOADING_ZHENGJIE_ZHU

ZHENGJIE ZHU

FORWARD DEPLOYED ENGINEER

I BUILD AGENT SYSTEMS AND STAY WITH THE CUSTOMERS WHO RUN THEM

$ whoami

I take agent systems from a customer's fuzzy problem to something running in their production environment — and stay on the hook for it.

$ cat ./approach.txt

  • SIT_WITH_THE_CUSTOMER/
  • SHIP_A_POC_FAST/
  • MAKE_IT_SURVIVE_PRODUCTION/

$ ./get_in_touch.sh

-- SECTION 01 ------------------------------------------------------------

SKILLS.EXE

DISCIPLINES AND TOOLS

  • [01]

    AGENTIC AI SOLUTIONS

    Own an engagement end to end — technical discovery, proof-of-concept, then a workflow the customer can actually deploy.

    STACK: PYTHON + MCP

  • [02]

    MULTI-AGENT ORCHESTRATION

    Peer-to-peer multi-round agent communication with team boundaries, audit trails, and bounded cost when rounds fail to converge.

    STACK: MATRIX PROTOCOL

  • [03]

    RETRIEVAL & RAG

    Hybrid vector and lexical search with reranking. Lifted Top-N recall from 65% to 85% on a production research corpus.

    STACK: MILVUS + ELASTICSEARCH

  • [04]

    RUNTIME & PLATFORM

    Configuration-driven onboarding for 200+ MCP services, plus elastic runtime startup work — shared caches and deferred init.

    STACK: DOCKER + K8S

  • [05]

    BACKEND & DATA INFRA

    Multi-level Redis caching, async RabbitMQ ingestion with dead-letter handling, Neo4j graph construction.

    STACK: POSTGRES + REDIS

  • [06]

    CUSTOMER ENGINEERING

    Technical point of contact for AI-native accounts — integration and inference debugging across model APIs, cache, and infra.

    STACK: WHATEVER BREAKS

-- SECTION 03 ------------------------------------------------------------

EXPERIENCE.LOG

WHERE THE WORK HAPPENED

  1. ALIBABA CLOUD NORTH AMERICA

    Sunnyvale, CA

    Solutions Architect & Software Engineer Intern

    Jun 2026 — Present
    • Co-architected agentic AI solutions with 3+ enterprise customers across hospitality, legal, and healthcare, owning each engagement from discovery through POC to a deployable workflow.
    • Technical point of contact for 5+ AI-native accounts; resolved 20+ integration and inference issues during 300% North America MaaS revenue growth.
    • Delivered a customer migration from Hermes to AgentRun so their agents run persistently on a sandbox-based runtime.
    • Built an internal MaaS analytics platform that cut weekly reporting from ~20 minutes to seconds.
  2. ALIBABA CLOUD — FUNCTIONAI (NOW AGENTRUN)

    Hangzhou, China

    Full-Stack Software Engineer Intern

    May 2025 — Aug 2025
    • Built a configuration-driven onboarding and deployment platform for 200+ MCP services, replacing per-service logic with reusable templates.
    • Sped up elastic runtime provisioning via reusable runtime layers, shared dependency caching, and deferred noncritical init.
    • Built a repository-aware frontend coding agent that turns design specs into production pages using existing components and design tokens — setup time down 80%, roughly two weeks to under two days.
  3. SEMANTIC COMPUTING LAB

    Jinan, China

    Research Software Engineer

    Oct 2023 — Dec 2024
    • Built a RAG workflow over Milvus vector search and Elasticsearch lexical search with reranking — Top-N recall 65% to 85%.
    • Designed a multi-level Redis cache with a customized LRU policy and TTL expiry, up to 3x API throughput in load testing.
    • Built async RabbitMQ ingestion for parsing, summarization, vectorization, and Neo4j graph construction, with retries and dead-letter handling.
    • Trained a multimodal hashing model with contrastive learning and knowledge distillation for cross-modal retrieval.

-- SECTION 04 ------------------------------------------------------------

ABOUT.TXT

WHO IS BEHIND THE SCREEN

$ cat bio.txt

I'm Zhengjie Zhu — a forward deployed engineer who likes the part of the job where you sit with a customer, find out what they actually need, and then go build it.

Most of my work is agent systems: multi-agent orchestration, retrieval pipelines, and the unglamorous runtime plumbing that decides whether any of it survives contact with production. I've shipped into a customer's production environment and stayed on the hook for it afterwards.

Currently finishing an M.S. in Computer Science at the University of Chicago. Before that, software engineering at Shandong University.

OPEN TO FORWARD DEPLOYED / SOLUTIONS ENGINEERING ROLES — SAY HELLO

  • 3+

    ENTERPRISE ENGAGEMENTS

  • 200+

    MCP SERVICES ONBOARDED

  • 85%

    TOP-N RETRIEVAL RECALL

  • 80%

    SETUP TIME REMOVED

$ cat education.txt

  • UNIVERSITY OF CHICAGOSep 2025 — Mar 2027

    M.S. Computer Science — GPA 3.6/4.0

  • SHANDONG UNIVERSITYSep 2020 — Jun 2024

    B.E. Software Engineering — GPA 3.9/4.0, Top 15%

$ ls ./tools — 6 ACTIVE

  • PYTHONagents, retrieval, servicesACTIVE
  • TYPESCRIPT / NODEfull-stack and toolingACTIVE
  • MCP + AGENT SKILLSagent loop engineeringACTIVE
  • MILVUS / ELASTICSEARCHhybrid retrievalACTIVE
  • POSTGRES / REDIS / NEO4Jstorage and cachingACTIVE
  • DOCKER / K8S / AWSdeployment and runtimeACTIVE

-- SECTION 05 ------------------------------------------------------------

CONTACT.SH

TELL ME WHAT YOU ARE BUILDING

CONTACT INFO

LOCATION
Chicago, IL
AVAILABILITY
Open to new roles

SYSTEM STATUS

  • AVAILABILITY100%
  • SHIPPING VELOCITY92%
  • REPLY SPEED90%
  • COFFEE LEVEL85%

$ compose_message.sh