AI software engineering platform engineer

Role Summary

Agent platform base

Build the Agent orchestration framework and runtime to support configuration expansion - adding/adjusting Agents can be completed by filling in forms or writing configurations, and try not to change the code.

• Build Agent registration and unified model gateway, shield underlying model differences, and support multi-model routing and grayscale
• Build an orchestration layer: allow multiple agents and tools to flow automatically in the process and reduce manual connection
• Measurement and Observability
• Build an Agent evaluation system: evaluation data set, regression baseline, quality indicator definition and automated evaluation pipeline
• Construction of online tracking and attribution: call links, failure mode classification, cost and token consumption management
• Make every prompt/model/arrangement change measurable, regressible, and rollable - Agents without evaluation are not allowed to enter production
• Context Layer

Open up the R&D tool chain (Git / Jenkins / Jira / Confluence / requirements and documentation system) and encapsulate the R&D context into consumable capabilities for Agents

Build unified context access based on MCP to handle engineering issues such as permission transparent transmission, incremental synchronization, cache isolation, and reference traceability.

Build enterprise research and development knowledge retrieval capabilities (RAG). Note: The difficulty in accessing knowledge is not in the retrieval algorithm, but in authority and timeliness.

Support teams in transitioning to AI Native

Establish Agent development specifications, scaffolding and templates so that domain engineers can deliver usable Agents without understanding the underlying LLM layer.

Support the transformation of engineers' capabilities: from "write it yourself" to "define the problem, hand it over to the agent, and verify the results"

Write documents, demos and training materials, organize practical exchanges, and lower the threshold for getting started.

Establish an adoption and effect measurement mechanism to transform feedback into platform iterations

Requirements

• Must
• Bachelor degree or above, computer/software engineering and other related majors
• Solid software engineering foundation, familiar with complete SDLC; proficient in Python (the team’s main language)
• Have experience in production and implementation of large-scale model applications - not just demo or POC, be able to explain clearly the problems encountered online and how to solve them
• Be familiar with the engineering implementation of Agent / RAG / vector database, and understand its failure modes instead of just happy paths
• Familiar with at least one Agent development framework: LangGraph / LangChain / OpenAI Agent SDK, etc.
• Familiar with R&D processes such as Git, CI/CD, Code Review, and automated testing

Other information

• Self-hosting: The platform is deployed on the corporate intranet and does not rely on public cloud hosting services.
• Connecting with existing systems: It needs to be integrated with existing internal tool chains, not reinvented
• Not tied to a single manufacturer: Do not lock in a specific chip/cloud/model manufacturer, the architecture must remain replaceable