Join Planera to build Manny, our AI scheduling assistant, and shape how construction schedulers work with AI on a modern Critical Path Method platform. You will own agent features end to end: designing and evolving the LangGraph/LangChain agent, engineering prompts and tools, integrating LLMs across providers, and holding response quality to a high bar with a real evaluation and observability stack. This is a hands-on applied AI role with a strong software engineering foundation and a focus on reliability, behavior quality, and user impact. You will work directly with the CTO and the lead AI engineer.
Key Responsibilities
•
Design, build, and own Manny features end to end across the agent backend, tools, and UI
•
Improve agent behavior, reliability, and answer quality through prompt engineering, tool design, and changes to the agent control flow
•
Evolve the agent architecture: ReAct loop, routing and controller logic, multi-node graphs, tool selection, and streaming responses
•
Integrate and tune LLMs across providers (Anthropic, OpenAI, Google), balancing quality, latency, and cost, including prompt caching and model selection
•
Design and extend Manny’s tool surface through the MCP server that connects the agent to Planera’s scheduling services
•
Build and own the evaluation loop: golden datasets, automated evaluators, snapshot-based replay, and offline and online quality metrics
•
Implement observability for agent runs with tracing, metrics, and structured logging, and use it to debug and improve behavior in production
•
Ensure safe, sandboxed execution of model-generated code and safe handling of tool side effects and mutations
•
Collaborate with product, backend, and frontend to deliver AI features end to end
Requirements
•
4+ years of software engineering experience, including recent hands-on work building production LLM features.
•
Strong proficiency in Python building production services
•
Hands-on experience building agentic systems with LLMs: tool and function calling, ReAct or similar loops, and orchestration frameworks such as LangChain/LangGraph
•
Practical prompt engineering skill: shaping model behavior reliably, debugging failures from traces, and managing large prompts and token cost
•
Experience evaluating LLM systems: building datasets, writing evaluators, catching regressions, and using tracing and observability tooling
•
Experience with the Model Context Protocol (MCP) or building tool and function-calling integrations for LLMs
•
Solid understanding of API design (REST, websockets, SSE and streaming) and interservice communication
•
Product mindset with a focus on user impact and pragmatic tradeoffs
•
Excellent remote communication skills
Preferred
•
Experience with MongoDB and Redis
•
Cloud experience (AWS or GCP), containers, and CI/CD
•
Go experience, as most of our backend systems are written in Go, including the MCP tool server
•
Practical experience with retrieval and augmentation (RAG), embeddings, and vector stores
•
Familiarity with LangSmith or comparable LLM evaluation and tracing platforms
•
Frontend or React familiarity for agent UI work
•
Domain knowledge in construction tech, project management, or scheduling