More jobs:
Job Description & How to Apply Below
Make eeze smarter and safer: retrieval, grounding and guardrails for an AI that negotiates with real money on the line.
eeze is an AI checkout. It answers buyer questions from a company's own docs, handles objections and can negotiate, but only inside rules the founder wrote, with every offer validated server-side. That combination of useful and safe is the whole product, and it is an engineering problem more than a prompt problem.
You will own the AI layer: retrieval quality, grounding and citations, evaluation, guardrails and cost. You will also help clients ship AI features in their own products during consulting engagements.
What you will do- Improve retrieval and grounding so answers cite the tenant's knowledge base accurately
- Design and enforce guardrails around offers, discounts and claims
- Build evaluation suites that catch regressions before customers see them
- Tune latency and cost across Anthropic and OpenAI models, including bring-your-own-key setups
- Turn knowledge-gap analytics into a product feature founders act on
- Prototype and ship AI features with clients during engagements
- Stay current on model capabilities and separate the useful from the noise
- Strong software engineering fundamentals; this is a production engineering role, not a research role
- Hands-on experience shipping LLM features to real users
- Practical knowledge of retrieval, embeddings, context management and evaluation
- Experience with at least one of the major model APIs in production
- A healthy scepticism and the habit of measuring model behaviour instead of assuming it
- Able to work from our Toronto office part of the week
- Experience with structured outputs, tool use and multi-step agent flows
- Background in payments, commerce or other domains where mistakes cost money
- Familiarity with guardrail and validation patterns for user-facing AI
- Experience running evals in CI
- You treat the model as a component, not a magic box
- You would rather ship a reliable small feature than demo an impressive fragile one
- You test with adversarial inputs because customers will
- You write down what you learn so the team compounds
- You care whether the answer is actually true
- 1 Intro call with an engineer, about 30 minutes
- 2 Short take-home assignment around retrieval and grounding on real docs
- 3 Technical conversation about your assignment and AI features you have shipped
- 4 Conversation with the founders
- 5 Offer
To View & Apply for jobs on this site that accept applications from your location or country, tap the button below to make a Search.
(If this job is in fact in your jurisdiction, then you may be using a Proxy or VPN to access this site, and to progress further, you should change your connectivity to another mobile device or PC).
(If this job is in fact in your jurisdiction, then you may be using a Proxy or VPN to access this site, and to progress further, you should change your connectivity to another mobile device or PC).
Search for further Jobs Here:
×