Staff AI Engineer - Agentic Shopping
Teampicnic · Amsterdam
What this role requires
7 requirements, read out of the advert rather than guessed from the job title:
Also mentioned, not required: English (B2). Worth having, but their absence is not what gets a CV filtered out.
See how often each of these is required across open data roles in Europe.
Free, no card. Tells you which of them your CV evidences and which it only implies.
Job description
In a nutshell
Build the intelligence that helps households get the week sorted
Agentic grocery assistance: Picnic brings together something unusually powerful for an AI builder: a direct customer relationship, a live grocery assortment and the operational system that takes every order from intention to delivery. That gives you the opportunity to build an assistant that does more than talk; it can understand a household’s plans and constraints, assemble a grounded basket, explain its choices and learn from what customers accept or change. You will decide where models, retrieval, ranking and deterministic services add real value, balancing customer control, correctness, latency, reliability and cost.
Technical leadership & vision: As a Staff AI Engineer, you will be the hands-on technical lead for Agentic Shopping. You will own technical direction, architecture and model-strategy decisions; solve the hardest production challenges yourself; and establish evaluation, observability, safety and fallback patterns that other teams can reuse.
Read the full description (33 more sections)Show less
Tricks of the trade
Build Picnic’s shopping assistant: Lead the technical direction for agentic shopping experiences that help millions of people plan and shop effortlessly.
Technical leadership & architecture: Lead the system architecture for multi-turn conversational agents, balancing latency, safety guardrails, and inference costs. Technically mentor engineers across the organization to execute on building production-grade agentic systems.
Rapid prototyping & productionization: Drive a fast-paced, iterative engineering cadence. Build initial prototypes, establish architectural patterns, and write production code when necessary to unblock complex technical challenges.
Model fine-tuning & context engineering: Design and implement systems architectures for multi-turn, conversational agents using open-weight models and techniques like fine-tuning, distillation, and RAG.
Evaluation & safety feedback loops: Establish continuous offline and online evaluation for mission completion, basket correctness, customer effort, clarification quality, acceptance and corrections, safety, reliability, latency and cost.
Cross-functional alignment: Collaborate with product, design, and product teams to define technical trade-offs, align on agent capabilities, and ensure a seamless, safe user experience.
We’re looking for
Proven track record of taking complex multi-turn conversational AI systems from architectural design to high-scale production.
Experience training, fine-tuning, and optimizing open-weight LLMs for production deployment.
Hands-on expertise building production RAG systems, context management, and state handling for multi-turn dialogues.
Experience designing robust evaluation loops, continuous monitoring systems, and automated safety/guardrail mechanisms for customer-facing applications.
Demonstrated ability to lead technical direction, perform design reviews, and mentor engineering teams without relying on hierarchy.
Strong Python skills and demonstrated ability to choose and productionize appropriate model, orchestration, retrieval, evaluation and inference tooling. Experience with tools such as PyTorch, LangGraph, vLLM, Pydantic AI, Qdrant or DeepEval is useful, or similar framework tools is required.
+ Bonus points!
Experience taking retrieval, ranking or recommendation models from problem definition to measurable production impact
In-depth expertise in deep learning for sequence modelling, representation learning, or retrieval and ranking
A degree in Computer Science, AI (or a related field), a PhD, open-source contributions, or peer-reviewed research are considered strong pluses
Picnic Perks
🫱🏼🫲🏾 Teamwork makes the dream work
With more than 80 nationalities across 3 countries, you’ll be part of a diverse company with plenty of cool stuff to get involved with, from board game evenings to after-work drinks to our company ski trip and more!
💡 Make a difference
You’ll work in an awesome startup environment with the freedom to drive your own projects and create a visible impact. Our fully electric vehicles and sustainable business model mean you’ll also be contributing to making the world a better place!
🥗 Fresh Lunch, coffee, and snacks
Our offices are equipped with fully-fledged coffee bars and a kitchen and chefs. They prepare delicious fresh and warm lunches every day so you can keep your energy up.
🚲 Health insurance discount & bike plan
We have a partnership with CZ (a health insurance provider). Picnic employees get a discount on CZ insurance packages between 5% and 15%. Furthermore, through our partnership with Lease a Bike, you can rent-to-own a new (e)bike at a discounted rate
🌎 Relocation
If you’re moving from another country to join Picnic we make it as smooth as possible for you. We’ll cover your flight costs for you and your partner and kids, your first month's rent and moving costs (up to €2000), and help you with the 30% tax ruling setup and application.
📆 All the rest
At Picnic you get 25 holidays, we cover your travel expenses and offer a pension plan. And your phone and laptop are on us, as well.
Find Jobs in Netherlands on Arbeitnow
You will apply. Then you will hear nothing.
And no one will tell you what was wrong. See it before you send: your ATS score, every weak line, and the fix for each.
- 1
Drop in your CV
One PDF, thirty seconds. No card.
- 2
See what is wrong with it
Every weak passage, quoted from your own CV, with the line to replace it.
- 3
Apply where you fit
Every European role ranked against what your CV actually says.
- Free, no card
- Your CV file is deleted after parsing
- Or skip the upload — build your profile by hand
- Refreshed every 6 hours