Forward Deployed Engineer
Levio
Posted May 15, 2026
- Other
- Données massives, intelligence artificielle et d'affaires / Big Data, BI & AI
Job description
### **Are you looking to thrive in a stimulating work environment?** Join Levio, a leader in digital transformation, and take your career to the next level. You will work alongside high-caliber professionals on ambitious, large-scale technology projects, directly embedded in our clients’ environments. At Levio, we value expertise, curiosity, and continuous improvement — and we give you the space to grow. ### About the Role We embed engineers directly inside enterprise clients to find, build, and ship AI and agentic systems that reach production and stay there. A Forward Deployed Engineer - FDE runs discovery with the customer's own stakeholders, scopes the use cases worth building, then writes and deploys the working system — in the customer's environment, against their data, under their security and governance constraints. This is a build-with-the-customer role, not an advisory one. You are the person in the codebase and in the room, closing the gap between an ambiguous "we should use AI" and a governed system operators actually trust and use. You own outcomes through real production usage — you do not hand off a slide deck or a prototype and move on. . ### Why Join Levio? - Work on complex, high impact digital transformation projects - Collaborate with experienced, multidisciplinary teams - Continuously develop your technical and professional expertise - Enjoy flexibility, autonomy, and a strong people first culture - Be part of an organization that values diversity, inclusion, and innovation ### Role and Responsibilities - Run discovery with the customer's business stakeholders and turn open-ended "we should use AI" asks into scoped use cases, ranked by ROI, data readiness, and technical feasibility. - Build and ship production GenAI and agentic systems end to end — retrieval (RAG), agent orchestration and tool/MCP integration, evaluation harnesses, and the surrounding application and data plumbing. - Deploy inside the customer's environment and constraints: their cloud, their security posture, their data governance — not a sandbox you control. - Design for trust — keep business-critical logic (pricing, rules, calculations) in deterministic code paths, build evaluation and human-in-the-loop review, and know when model output is safe to put in front of operators. - Embed and enable: document the system, run walkthroughs, and lift adoption — while owning the system through real production use, not until the demo. - Present architecture, cost, and delivery trade-offs directly to non-technical decision-makers and keep them bought in. ### Qualifications and Experience - Bachelor’s or Master’s degree in Computer Science, Engineering, Data Science, or related field - 6+ years of experience in engineering roles, with 3+ years in GenAI/Agentic Systems positions - Strong communication skills and the ability to explain technical concepts and trade-offs to both technical and non-technical audiences. - Proven ownership of systems through production deployment, adoption, and iteration. ### What we require - **You ship code now.** Recent, hands-on, first-person production work you can walk through line by line — not architecture you directed or reviewed. If your last real commit was two years ago, this isn't the role. - **Real engineering fundamentals.** You can drop into an unfamiliar codebase, debug code you didn't write, and build the thing when the platform's GUI can't express it and the AI assistant is stuck. Low-code and AI-assisted development are welcome as tools, not as the whole toolbox. - **Genuine customer-facing experience.** You've been embedded with customers — in their environment, their mess, their competing priorities — and navigated ambiguity and stakeholders with no direct authority. Internal-only product work doesn't demonstrate this. - **Demonstrable production GenAI depth.** You've built real RAG and/or agentic systems and can explain the mechanics — retrieval strategy, orchestration, tool-calling — and, above all, how you evaluated them and made them safe to trust. A buzzword list is not depth. - **Evaluation and guardrail instinct.** You keep the stochastic model out of business-critical decisions, and you can prove your AI output is trustworthy rather than hoping it is. - Code we can look at and claims you can substantiate in a live session. A résumé is a starting point, not evidence. - **Ownership temperament.** You can own a system through production and iteration, inside a team's codebase and discipline — not only build-and-leave proofs of concept. ### Strong pluses - Regulated-industry delivery (financial services, insurance, actuarial, wealth, pension) and comfort with its data, privacy, and governance realities — RBAC, audit, PIPEDA/GDPR. - Governed enterprise-AI patterns: human-in-the-loop, confidence thresholds, audit logging, escalation-path design. - Working depth in one or more of LangGraph or comparable agent frameworks, MCP, and AWS Bedrock / Azure OpenAI / GCP — real depth in one beats a shallow tour of all three. - A track record of taking use cases from discovery to production with measurable, defensible outcomes. - Strong written communication and the ability to teach — enablement is part of the job. ### What this role is not - **Not an architect or advisory seat.** You build and deploy; you don't just diagram, cost, and present. - **Not a low-code or platform-configuration role.** Platforms are means; the job is engineering. - **Not prototype-and-handoff.** You own outcomes through production, not until the demo works. - **Not a solo BD or fractional-consulting arrangement.** This is full-contact delivery on a team. ### How we evaluate We weight what you can demonstrate over what's on the page. Come ready to build, and bring specifics. - **A live, assistant-off exercise** — extend or debug an unfamiliar codebase in front of us. This is the primary signal. - **A production deep-dive** — walk us through what broke in production, how you found it, and how you knew your AI output was safe to trust. - **A discovery role-play** — turn an ambiguous stakeholder ask into a scoped, ROI-ranked use case, out loud. - **A depth check on your own claims** — pick your strongest AI project and go as deep as we push: the graph, the retrieval, the evals, the guardrails. ### Fit check **You'll thrive here if:** - You'd rather ship a working system in a customer's messy environment than produce the perfect architecture diagram. - You're energized by ambiguity and being the most accountable person in the room. - You treat "the model said so" as the start of the work, not the end. **This isn't for you if:** - You've moved into full-time architecture, strategy, or team leadership and don't miss the code. - Your AI experience is mostly demos, courses, or platform configuration. - You want to advise and hand off rather than build and own. ###