AI Engineer

Every enquiry, application and enrolment runs through Salesforce, and every family finds us first through efacademy.org.

We want one engineer who builds with AI at the frontier and makes the rest of the organization better at it. You will ship agents and AI workflows into production, prove they work, and turn what works into practices that a small engineering team and a much larger group of non-technical colleagues use every day.

This is a hands-on engineering role, not a training or strategy role. You will write code, ship agents into production and own the results.

What success looks like

• Within 30 to 60 days: our Agentforce agent is live in production, answering families’ questions with an eval set behind it and results we measure every week.
• By day 90: at least one more AI workflow in production that colleagues use weekly, with a measured before and after.
• By month six: a written, practiced standard for how we build with AI, and AI workflows running in admissions, marketing or finance with measured results.
• By the end of year one: business teams bring you problems, not tool requests, and people outside technology build their own simple automations safely.

How we hire 

1. Application. Send your CV and answer one question in a few paragraphs: Describe something you built with AI that changed how you or your team works. What was it like before, what changed, and how did you know?
2. Interview with the VP of Technology.
3. Work sample. Up to three hours on a small Salesforce project we built for hiring: find and fix two reported problems, then build an action for an AI agent. Use any AI tools you want, and send your session transcripts with the result.
4. Technical interview with the VP of Technology and a member of our technology team. You walk us through your work sample and we ask you to explain any line of it. We assess how you worked as much as what you produced.
5. Final conversation with the Executive Vice President of Sales and Marketing, covering culture and motivation.

We ask every candidate the same core questions, we decide quickly, and you will hear back from us either way.

Practical details

• On-site: five days a week in our Zurich office.
• Employment: permanent EF employee, not a contractor.
• Visa: we can support work permit sponsorship for the right candidate.

Required

• 5+ years of professional software engineering, with production code you wrote, reviewed and maintained after it shipped. Git, pull requests, CI/CD and tests that catch regressions are routine for you.
• Hands-on Salesforce development: Apex, Lightning Web Components, SOQL and Flow, with Salesforce DX, plus integration through REST APIs and event-driven patterns such as Platform Events.
• 2+ years shipping LLM-powered features or agents to real users, not prototypes. Show us one that people still use and what it changed.
• Agent engineering: you have designed tools, structured outputs and multi-step agent loops, and you know how they fail. Hands-on experience with the Model Context Protocol (MCP).
• Evals: you have built eval sets from real tasks, run them automatically, and used the results to decide whether a prompt, model or code change ships. You know where code checks work and where model-graded evals are needed, and the limits of each.
• LLM security and privacy: defenses against direct and indirect prompt injection, personal data leakage and over-permissioned tools, including when a human must approve an agent’s action.
• Production trade-offs: model choice, prompt caching, batching, retries and fallbacks to balance quality, cost and latency.
• Daily use of coding agents, such as Claude Code, which we use. You read, test and own everything an agent writes for you.
• Judgment about where AI is the wrong answer.
• Clear writing, and the patience to help non-technical colleagues change how they work.
• Professional English.

Strongly preferred

• Agentforce in production, or Einstein and Data Cloud.
• Strong TypeScript or Python for work outside Salesforce, such as MCP servers and eval tooling.

Nice to have

• Our web stack: React and Next.js, and AWS serverless (Lambda, API Gateway, S3, EventBridge).

• Salesforce Platform Developer I or II, or Agentforce Specialist certification.

• Retrieval: hybrid search, reranking and vector stores.

• LLM observability tools, such as Langfuse, LangSmith or OpenTelemetry tracing.

• Agent SDKs or frameworks, such as the Claude Agent SDK, OpenAI Agents SDK or LangGraph.

• Infrastructure as code, ideally Terraform.

• Experience Cloud portals, or a headless CMS such as Storyblok, plus Tailwind and Vercel.

• Public work: open source, writing or talks about what you built with AI.

If you match most of this but not all of it, apply anyway.

• Agents in Salesforce. Build and run agents that answer family questions and nurture leads across SMS, WhatsApp and web chat, on Agentforce or alongside it where another tool does the job better.
• Evals and monitoring. For every agent and workflow you ship, build the eval set from real tasks, run it on every change, and track quality, cost and latency in production, including after model updates.
• AI in the engineering workflow. Set how our small technology team uses coding agents for implementation, tests, review and documentation. Build the shared tooling that makes it work, such as MCP servers for our systems, reusable prompts and CI checks, and measure whether it helps.
• AI for the business. Find repetitive work in admissions, enrollment, marketing and finance that an agent or automation could do. Build it, ship it and measure the result.
• Guardrails. We hold personal data of students, many of them minors, under GDPR and Swiss data protection law. Decide with us what AI may touch, and design every system against prompt injection, data leakage and agents acting beyond their remit.
• Teaching by building. Write the standards, pair with our developers, and help non-technical colleagues build simple automations safely.
• Staying current. Track new models, tools and techniques, test them on our own work, and recommend what we adopt based on evidence.