高级软件工程师 II（AI赋能）
Senior Software Engineer II (AI Enablement)
Aledade的AI赋能团队构建并维护其余AI应用所依赖的平台。AI赋能平台工程师是负责基础架构的工程师：MCP网关，它处理代理与Aledade系统之间的所有工具调用；插件市场和安装器，将代理能力分发给工程师和非工程师；模型网关配置，决定Aledade在推理上的支出及表现；以及使所有内容可衡量的遥测系统。
这是一个具有异常直接业务影响的平台角色。该工程师负责的模型路由和测试评估工作，是区分多元化、成本优化推理策略与单一供应商策略的关键。这是适合希望获得平台深度而无需平台抽象的工程师的职位：真实用户、真实支出、可衡量的采用率，以及设计决策与其影响之间的短距离联系。
主要职责：
- 负责并扩展MCP网关和连接器平台。在网关上构建并强化MCP服务器集成（Glean、Slack、Jira、Snowflake、Salesforce、Monday、Tableau、Databricks等）；负责非BAA工具的认证、作用域、租户、速率限制和读写策略执行；与安全团队和平台负责人合作，建立默认安全的访问路径。
- 负责分发层，运行市场发布流水线：版本控制、CI/CD、插件PR评审、发布卫生和开发者工具同步流程。构建和维护被PTA团队广泛使用的共享插件、技能、钩子和基于规范的开发工具，以及让其他团队能够自主贡献的作者指南。
- 负责平台可观测性、模型网关配置和成本归因。扩展使用汇总和遥测流水线，使采用率、工具调用健康状况和推理支出可以按团队、工作负载和模型进行归因。定义模型路由并构建基准测试和测试/评估，使Aledade能够基于证据选择模型。
- 扩展代理输出和协作界面。构建基础设施以支持将代理输出映射到可浏览、可搜索、可评论的界面，包括发现和标签、每观看者的参与度监控、评论、过期和分享通知。
- 赋能和平台管理。作为实践者开展问答时间及午餐交流；指导工程师掌握代理编程和插件编写模式；处理并分析问题。
查看英文原文
Aledade's AI Enablement team builds and runs the platform that the rest of Aledade's AI adoption depends on. The AI Enablement Platform Engineer is the engineer accountable for the substrate: the MCP Gateway that brokers every tool call between agents and Aledade's systems, the plugin marketplace and installer that distribute agentic capability to engineers and non-engineers alike, the model gateway configuration that determines what Aledade spends on inference and how well it performs, and the telemetry that makes all of it measurable.
This is a platform role with unusually direct business consequences. The model-routing and harness-evaluation work this engineer owns is what distinguishes a diversified, cost-optimized inference strategy from a single-vendor one. It is the right role for an engineer who wants platform depth without platform abstraction: real users, real spend, measurable adoption, and a short line between a design decision and its effect.
Primary Duties:
- Own and extend the MCP Gateway and connector platform. Build and harden MCP server integrations (Glean, Slack, Jira, Snowflake, Salesforce, Monday, Tableau, Databricks and beyond) on the gateway; own auth, scopes, tenancy, rate limiting, and read/write policy enforcement for non-BAA tools; partner with Security and platform owners on safe-by-default access paths.
- Own the distribution layer, run the marketplace publishing pipeline: versioning, CI/CD, plugin PR review, release hygiene, and the developer tooling sync processes. Build and maintain shared plugins, skills, hooks, and Spec-Driven Development tooling consumed across PTA teams, plus the authorship guides that let other teams contribute without hand-holding.
- Own platform observability, model-gateway configuration, and cost attribution. Extend the usage-rollup and telemetry pipeline so adoption, tool-call health, and inference spend are attributable by team, workload, and model. Define model routing and build the benchmarks and harness/evaluations that let Aledade choose models on evidence.
- Extend the agent output and collaboration surface. Build out infrastructure to support mapping agent outputs into a browsable, searchable, commentable surface, discovery and tagging, per-viewer engagement instrumentation, comments, expiry, and share notifications.
- Enablement and platform stewardship. Run office hours and brownbags as a practitioner; mentor engineers on agentic-coding and plugin-authorship patterns; triage and dedupe inbound feature requests; carry on-call for marketplace-published artifacts and gateway availability.
Minimum Qualifications:
- BS/BTech (or higher) in Computer Science, Engineering or a related field.
- 5+ years professional software engineering experience.
- Production ownership of a backend service or developer platform. API gateway, service proxy, SDK/CLI, internal developer platform, or comparable, including its auth model, release process, and operational health.
- Strong production experience in at least one modern application stack (Python, TypeScript/Node, Go, or similar) and modern CI/CD.
- Hands-on experience with authentication and authorization for machine-to-machine traffic (OAuth2, OIDC, M2M credentials, token scoping, secret management).
- Demonstrated experience instrumenting a system you own; metrics, structured logging, tracing, or usage analytics, and using that data to drive a decision.
- Direct hands-on experience with one or more agentic coding tools in a production or near-production setting (Claude Code, Cursor, Cody, Copilot agents, Aider, or equivalent).
- Strong written communication: comfortable producing documentation, runbooks, and educational artifacts for engineers who weren't in the room.
Preferred KSAs:
- Experience authoring or maintaining MCP (Model Context Protocol) servers, Claude Code plugins, skills, hooks, or comparable LLM-tooling integrations.
- Experience running an LLM gateway or inference proxy in production (LiteLLM, Bedrock, vLLM, or similar) — routing, fallback, caching, quota, and cost attribution.
- Experience building evaluation harnesses or benchmarks for LLM systems, including A/B comparison of prompts, tools, or model versions.
- Experience operating a package registry, plugin ecosystem, or extension marketplace — publishing pipelines, semantic versioning, compatibility, and deprecation.
- Background in healthcare technology, HIPAA-regulated environments, or PHI-handling systems; familiarity with BAA and data-residency constraints on third-party tooling.
- Familiarity with Aledade's stack (Python/FastAPI, Vue/TypeScript, Postgres, AWS, Auth0, Datadog, Sumo Logic) is a plus but not required.
- Comfort across the full stack (frontend/backend/infra) — platform work at this stage doesn't honor team boundaries.
- Experience as an early engineer on a platform whose users are internal colleagues, where adoption has to be earned rather than mandated.
Physical Requirements:
Sitting for prolonged periods of time. Extensive use of computers and keyboard. Occasional walking and lifting may be required.