SlipstreamJobs tracks this role from the company's public career site. Apply directly on the employer's site.
Cambio is a Y Combinator-backed software platform for real estate decarbonization, helping commercial real estate owners achieve net-zero portfolios through an AI-driven sustainability, asset, and investment management platform. The company is backed by Google and Stanford and is the fastest-growing startup in its category.
You will build AI-powered features end-to-end, from prototype to production. Your responsibilities include designing and running evaluations to measure and improve model and feature quality, building and maintaining data and inference pipelines that power AI features in production, experimenting with new models, prompting strategies, and retrieval approaches to improve accuracy and cost-effectiveness, and supporting and collaborating with teammates.
The role emphasizes shipping through obstacles, moving with urgency, and closing each week with visible impact. You will own your work end-to-end, proactively manage visibility, and over-communicate to ensure alignment. You are expected to show up for teammates in person, unblock issues before being asked, and never let someone stay stuck.
Cambio values A-players with low ego and genuine curiosity about their craft. The culture prioritizes outcome-driven work, iteration speed as a quality metric, and the belief that building fast and learning faster leads to great products. Work is viewed as part of life because team members genuinely enjoy it.
The company offers offices in Waterloo, New York, and San Francisco, with the expectation of in-person presence.
REQUIREMENTS:
- Experienced in Python web development (Django is a bonus)
- Shipped AI features and understand evaluation challenges in production
- Hands-on experience with LLMs/generative AI (prompting, fine-tuning, or RAG)
- Comfortable with evaluation frameworks and measuring model quality in production
- Familiar with vector databases and retrieval-based architectures
- Understands tradeoffs between model cost, latency, and accuracy
