Principal Full-Stack AI Engineer
This position is hosted under the Constellations Fellowship Program at Climate Curve.
This summer, we are offering 50+ part-time, fully remote fellowship opportunities with 20+ KCP Laureate and Finalist organizations working at the forefront of climate innovation. These fellowships are designed for students and recent graduates seeking hands-on experience and meaningful careers in the climate space.
How to Apply:
- Review available fellowship positions, including responsibilities, qualifications, and time commitments:
https://www.climatecurve.org/constellations-positions - Submit your application here:
https://www.emailmeform.com/builder/form/P90wdZ81aXU7S9xz0CHjvQl
Application Deadline:
Sunday, May 3rd at 11:59 PM EST
Principal Full-Stack AI Engineer Fellow
About the Organization:
Climatize is building the AI system that finds the best renewable energy sites, designs optimal projects, and funds them. We compress what takes developers weeks of manual diligence into minutes, while deploying $250K to $5M in project debt that banks can't efficiently fund.
Our AI prospecting engine analyzes grid capacity, incentives, zoning, and interconnection data to surface high-value sites.
Our execution engine then automates underwriting, designs the project, and assembles investor-ready documentation.
Fellowship Description:
This is a building role, not a planning role. You will pair directly with the Director of AI to ship features across the full platform. The core of this role has shifted: AI agent systems are now primary output. You will own agent pipeline architecture end-to-end using Python, CrewAI, LangGraph, and LangChain alongside the TypeScript full-stack.
Day-to-day work spans: agentic workflow design, React components, Playwright automation, PostGIS queries, and external API integrations. Claude Code is used extensively and fluency with AI-assisted development is a real multiplier here.
What You'll Do:
•Design and ship multi-agent pipelines using CrewAI, LangGraph, and LangChain for solar project automation
•Build LangGraph state machines that orchestrate the 9-phase HelioScope design pipeline end-to-end
•Implement CrewAI agent crews for document extraction, permit research, and feasibility analysis
•Instrument LangSmith traces for agent observability, failure replay, and latency optimization
•Integrate OpenAI / Anthropic APIs with structured outputs (function calling, tool use, JSON mode)
•Prompt engineering with chain-of-thought, few-shot, and RAG patterns against domain corpora
•React + TypeScript SPA (shadcn/ui, Tailwind CSS, React Hook Form, Zod)
•Interactive maps (Mapbox GL), vector tile overlays (HCA hosting capacity), polygon drawing tools
•Real-time agent progress via SSE streams from LangGraph event APIs
•Data visualization: financial models (Recharts), load profiles, solar production forecasts
•Playwright automation pipeline driving HelioScope solar designs (production, flaky DOM tolerance)
•Google Solar API integration (building masks, roof segments, irradiance data)
•Geometry transformations: EPSG:4326/3857 conversions, polygon simplification, fire code setbacks
•New external data sources: utility APIs, ArcGIS endpoints, OpenDataSoft
•Python FastAPI or Node.js Express services backing agent workflows and permitting pipelines
•PostgreSQL schema extensions and Supabase integration for multi-tenant project data
•PostGIS spatial queries for vector tile serving and HCA data
•Vector embeddings (pgvector / Pinecone) for RAG over solar interconnection documents
•Unit and integration tests for agent steps (pytest + LangSmith evals)
•CI/QA cycles for the DDR automation pipeline; maintain coverage as codebase scales
Qualifications:
•5-7+ years shipping production software; at least 3 in React + TypeScript owning features end-to-end
•Hands-on Python proficiency: asyncio, Pydantic models, FastAPI or equivalent; this is not optional
•Demonstrated experience building with LangChain or LangGraph: chains, agents, tool calling, state machines
•Production CrewAI or equivalent multi-agent framework experience: crew design, role definition, task routing
•LLM API fluency: structured outputs, function calling, context window management, cost and latency tradeoffs
•Strong Node.js backend: Express APIs, async patterns, SSE/streaming, messy third-party integrations
•PostgreSQL depth: migrations, query optimization, JSONB, PostGIS spatial types
•Demonstrated autonomy: you break it down, ship it, and flag risks without being asked
Tech Stack:
AI Frameworks
CrewAI · LangChain · LangGraph · LangSmithLLM
APIs
Anthropic Claude · OpenAI · Gemini
Python
FastAPI · Pydantic · asyncio · pytest
Frontend
React · TypeScript · Tailwind · shadcn/ui · Mapbox GL
Orchestration
Playwright · Node.js · SSE Streams
Data
PostgreSQL · PostGIS · pgvector · Supabase · Pinecone
Geospatial
EPSG transforms · Turf.js · ArcGIS · Google Solar API
Infra / DX
Claude Code · Cursor · GitHub Actions · Docker
Time Commitment: 30 hours / week