AI Research Engineer (Tool Calling Safety)
Share
Job Location:
Monthly Salary:
Posted on:
30+ days ago
Vacancies:
1 Vacancy
Department:
Job Summary
About StackOne:
StackOne is the AI Integration Gateway for SaaS products and AI Agents. Backed by GV and Workday Ventures ($24M raised) we help builders of SaaS platforms and AI Agents orchestrate hundreds of scalable accurate and enterprise-grade integrations. Our platform combines 25000 pre-mapped actions on 200 connectors an AI-powered integration development toolkit plus security by design: a real-time architecture managed authentication and permissions and end-to-end observability.
Join us on our fast trajectory to build the future of agentic integrations.
With an AI-native integration toolkit that delivers real-time execution managed authentication granular permissions and full observability all built with safety at its core were now doubling down on AI R&D & creating our own lab to push the boundaries of tool calling for agents: training specialized LLMs designed to outperform general-purpose models in what matters most: precision reliability and safety in agentic execution.
About the role
Youll help build a world where users of any agents can integrate with the tool of their choice in one click thanks to StackOne.
Were looking for an AI Research Engineer to secure the next generation of tool-calling for LLMs. Youll research and prototype defences that limit the attack surface of emerging standards such as MCP servers and make tool execution safer against prompt injection and malicious calls.
Youll work with our researchers on state-of-the-art tool-calling models and with our engineers on the execution engine that runs them.
You will own design and run experiments on cutting-edge architectures manage distributed training clusters and help curate & generate high-quality datasets. This role sits closer to the research/ML infra side than product engineering but with a strong mandate for applied production-ready results.
In this role you will work with wider AI team of StackOne (comprising of other researchers and engineers) and report directly to the CTO.
Responsibilities
Investigate vulnerabilities in tool-calling protocols (MCP OpenAPI custom plug-ins).
Design and test mechanisms to restrict malicious or unsafe tool calls.
Collaborate on the architecture of a secure tool execution engine.
Partner with the core research team to align safety work with model quality.
Design and manage synthetic data generation workflows to augment real-world datasets.
Develop tools for dataset curation labeling filtering and augmentation.
Conduct benchmarking and evaluations to measure fine-tuning impact.
Collaborate with the rest of the engineering team to integrate fine-tuned models into production stacks.
Stay ahead of research in parameter-efficient fine-tuning synthetic data and LLM training.
What were looking for
Knowledge of prompt injection sandboxing and least-privilege patterns.
Experience building or auditing APIs plug-ins or protocol runtimes.
Ability to design and evaluate defences for real-time agent systems.
Background in Gen AI with experience with Tool Calling
Interest in security engineering or applied cryptography.
Desire to work in a fast-paced startup taking ownership of projects e2e and bias towards shipping.
Benefits
25 days holiday 1 additional day holiday per year of tenure
Participation in the companys employee share options plan
Private health insurance (including dental & optical)
Health fitness and gift card discounts
1000 for your home office set up 500/year top-up
Paid lunch in the office
Annual team offsite to sunny spots (last ones were in Spain and Portugal )
Join one of Europes fastest-growing startups
Work with a veteran team of ex-employees of Google Microsoft Oracle Coinbase JP Morgan and more
Cycle2Work and Electric Cars scheme
Hybrid work set up - typically 2d in the office
Ready to help us change the game for AI Agents Get in touch and lets chat!
We believe diversity drives innovation. We encourage individuals from all backgrounds to apply. As an equal-opportunity employer we celebrate diversity and are committed to creating an inclusive environment for all employees.
Key Skills
- ADAS
- System Design
- Distributed Control Systems
- Technical Writing
- Systems Engineering
- Requirements Management
- OSHA
- Sensors
- Signal Processing
- Microsoft Project
- Budgeting
- Programmable Logic Controllers
About Company
One integration platform, two powerful interfaces: Unified APIs for SaaS products, and AI Agent Actions. Ship hundreds of integrations in days.
