Cracked AI Engineer

AI Startup · Toronto, Canada

About Us

We're solving one of the largest problems in the physical world: housing that's too slow and too expensive to build. The $2 trillion construction industry runs on human expertise that doesn't scale - so we're building the AI that understands what to build and how to build it.

We start with the costliest problems for general contractors - the multi-million-dollar decisions that Claude cannot solve. Tyce Agent wins more projects and estimates costs on multi-million-dollar builds live at one of the landmark construction companies today.

We just closed a 7-figure round and are hiring to build the best AI platform for construction.

Competitive salary and equity.

What You'll Do

You'll own the AI core of the product - how we ingest, understand, retrieve, and reason over customers' documents. This is the heart of our moat, and it's where the hardest, most interesting problems live.

· Building a Context/Knowledge Graph around our customers knowledge bases - process messy PDFs, Excels into AI-native context representation

· Improve on our agent harness - agentic search and any unique tools as part of our platform

· Build the evaluation harness that keeps quality high as we ship fast.

· Ship latest AI featureas day-to-day alongside the founders and shape the AI roadmap.

We're on a modern, lean stack: FastAPI, PostgreSQL, Pydantic-AI, Logfire, AWS, with a React Router / TypeScript frontend.

Required Qualifications

· Mid-to-senior level of AI/engineering experience.

· Strong in Python; comfortable shipping production systems, not just notebooks.

· Hands-on experience building and evaluating LLM/RAG or ML systems.

· Go-getter mindset, personal agency, fast learner, excellent verbal and writing skills.

Nice-to-have Qualifications

Having any of these is a strong plus — tell us your superpower:

· Built and evaluated an AI agent / RAG system for real-world performance.

· Document parsing / OCR / information-extraction experience (PDF, DOCX, tables, layout).

· Retrieval/search background (embeddings, ranking, vector + keyword hybrid).

· Comfortable full-stack enough to ship your own features end-to-end.

· Competitive programmer (e.g. ACM ICPC).

Check us out at https://www.tyce.ai/