Senior Applied AI Engineer – Agent Runtime (Alwin by DataSnipper)
About this role
The day-to-day work for a Senior Applied AI Engineer on the Agent Runtime team involves owning the core components that enable Alwin, DataSnipper's Agentic Automation Platform. This platform is designed for audit and finance, where agents execute complex, multi-step procedures over client evidence, ensuring every output is traceable back to its source, with final sign-off from humans. This role sits at the foundation of this system, focusing on the applied AI aspects of the runtime layer, including model access, system prompts, tool integration, agent orchestration, and performance measurement.
This position is critical because product teams build domain-specific agents on top of this runtime, and this engineer determines how the base agents reason, what tools they utilize, and how context is managed over extended execution periods. A key function is establishing the production standards for agents in a regulated domain, as there is currently no established playbook for this type of deployment.
This role is ideal for an engineer who thrives on solving open-ended production problems and enjoys bridging the gap between complex AI concepts and reliable, scalable software systems. Individuals who are detail-oriented, capable of defining what 'correct' means in a regulated context, and driven to improve agent performance through rigorous evaluation would perform well in this environment.
What you will do
- Own the core components of the base agents, including system prompts, context engineering, and memory management for tasks spanning many turns and hours of execution.
- Design the methodology for how agents interact with the tool layer, determining the appropriate level of abstraction for document extraction, retrieval, and sandbox usage.
- Implement advanced agentic patterns such as sub-agent composition, planning modes, human-in-the-loop gates, and model routing across different providers.
- Manage the end-to-end delivery process, taking agents from initial prototypes with domain experts through evaluation to final production deployment.
- Develop and build automated processes designed to continuously improve agent performance over time.
- Extract meaningful signals from long agent trajectories to attribute outcomes to specific reasoning steps and failures, converting these into actionable fixes.
- Measure accuracy, latency, and token cost, explicitly defining trade-offs for downstream teams building on the runtime.
- Instrument agent behavior with the observability stack to leverage production traces for continuous improvement.
What you need
- Candidates must possess at least five years of software engineering experience, supported by strong, production-grade Python skills.
- Experience shipping and operating an LLM-powered product in production is required, specifically handling issues like hallucinations, latency spikes, tool failures, and cost overruns at scale.
- Hands-on experience building agentic systems, including control loops, tool selection logic, planning versus execution strategies, and implementing retries and fallbacks beyond simple prompt-and-parse pipelines.
- Proficiency is needed in evaluating AI systems, demonstrated by building datasets, running offline and online experiments, and using results to measurably enhance an AI system.
- Fluency with LLM APIs and agent frameworks across multiple model providers is necessary.
- Strong fundamentals in software architecture and system design, coupled with a proven track record of delivering reliable and well-tested software.
- Excellent communication skills are required to collaborate directly with auditors and product partners to establish criteria for correctness.
Nice to have
- Experience with Durable workflow engines or long-running background task systems
- Experience with RAG and retrieval pipelines over large, messy document corpora
- Document AI experience, including VLMs, OCR, structured extraction, and their metrics
- Experience with sandboxed code execution, MCP, or multi-agent orchestration in production
- Domain experience in audit, accounting, or fintech
- Familiarity with OWASP GenAI security practices and working in regulated, privacy-sensitive environments
Benefits
- Headquartered in Amsterdam with offices in New York, Tokyo, Kuala Lumpur and Sydney
Skills and tech stack
Questions candidates ask
What is the remote policy for this role?
The posting does not explicitly detail the remote work policy. The company is headquartered in Amsterdam with offices in New York, Tokyo, Kuala Lumpur, and Sydney.
What is the compensation structure?
The posting does not specify salary details for this position.
What is the primary technology stack used?
The core technologies mentioned include Python, LLM APIs, and various agent frameworks. The runtime layer involves components like document extraction, retrieval, and sandboxes.
What is the team like?
This is a hands-on role within a small team that has a large blast radius. The team partners with Agent Experience teams, Document Intelligence, and the AI Platform team.
What is the growth trajectory for this role?
The role focuses on helping to write the playbook for production agents in a regulated domain, indicating a focus on establishing new standards and processes.