Full Stack Engineer (AI and agentic workflows)
Job description
About i23
i23 is building a generational AI-native commercial insurance broker, starting in the UK.
Insurance broking is a vast and highly cash generative business. Brokers bring in an estimated $700 billion of revenues every year. They have been around for hundreds of years. They still run on legacy systems and processes.
AI has created a structural opportunity in the sector by enabling agentic operations that incumbents will be too slow to replicate. The winners of tomorrow will be different from the leaders today. Our opportunity is to be the first, build the brand and take the category.
Brokers and their teams still spend too much of their time gathering information and chasing paperwork. i23 is developing an agentic operating model that takes on that routine work and helps brokers grow revenue, while people stay responsible for advice, professional judgement and client relationships. Every renewal, quote and claim will feed back into i23’s learning systems, accelerating compounding growth with every policy it binds.
i23 was founded by Anand Sambasivan (CEO), co-founder of PrimaryBid (Series C scaled fintech, backed by SoftBank and sold to SoFi), and Bim Afolami (President), former City Minister and Economic Secretary to the Treasury, and most recently Head of Strategy & Execution at Acrisure.
Engineering is led by Dr. Salman Valibeik, who holds a PhD in AI, computer vision and robotics from Imperial College London and previously co-founded and led ORPIVA, a generative AI company, as CEO.
The company has raised an oversubscribed $5.5m seed round from Outward VC, Seedcamp, Insiders VC and Four Cities Capital, alongside strategic angel investors, as covered recently in Sky News.
The role
As a founding engineer, you’ll be responsible for rapidly shipping agentic workflows to the brokers serving i23’s clients. If you execute well, your work will show immediate impact on our KPIs. You will have access to insurance data, experts, and live users.
The work sits at the intersection of applied AI and insurance expertise, where the technical problems are hard and largely unsolved. Winning here will mean building systems that support continual learning, and investing in high-fidelity synthetic data, rigorous testing, and agent simulation infrastructure to measure performance across a carefully curated distribution of tasks. Because insurance is regulated, it will also be critical to design for reliability and to thoughtfully balance AI automation with human judgement.
The role is highly dynamic and cross-functional. We’re looking for a high ownership generalist who’s hungry to learn, build and test. One day you may be interviewing a leading UK expert on property insurance, the next you’ll be encoding expertise into an agent that helps brokers win 10 times more business. The right founding engineer will bring strong opinions on where AI is heading, along with the judgement to decide which parts of the stack to buy versus build and where i23 can create a durable moat as an AI-native broker.
You’ll work directly with i23’s founding team. As part of a small team, you’ll help shape the company’s strategy, brand and technical roadmap as well as the code.
What You Bring
- You’ve built and maintained full-stack applications end-to-end, from data models and pipelines to the user interface, and run them in production with real users.
- You’ve built agentic workflows: agents that plan, call tools, recover from errors and hand off to humans.
- You’re comfortable researching, trialling and picking up new tools and languages quickly.
- You can take a vague business problem and turn it into a build plan.
- You explain technical work clearly and can map technical tradeoffs to business impact.
- You take sensitive data seriously, and treat access control, encryption and audit trails as part of the design rather than an afterthought.
- You’re a curious person, and your interests extend beyond software.
- You have the right to work in the UK.
Bonus Points If You Have Experience With
- Pydantic AI, Google’s Agent Development Kit, LangGraph or another agent framework.
- Temporal, DBOS or another durable workflow engine.
- Document parsing: PDFs, scanned forms and tables.
- Classic machine learning, such as forecasting or classification, alongside LLMs.
- Terraform or other infrastructure as code.
- LLM tracing and evaluation tools such as Langfuse.
- Building custom evaluation pipelines and automated benchmarking for complex LLM workflows.
- Defending AI systems against prompt injection.
- Worked in a regulated industry.
- Time at an early-stage startup.
What We Offer
- A founding role with real ownership of the product and the technical direction.
- Your work in front of real users quickly.
- A team of experienced engineering and product leads to learn from and build with.
- A competitive salary.
- An office in Temple, central London.
- Holiday, pension and other benefits.