hey, I'm erikrasin.io đź‘‹
Erik Rasin
AI builder for high-stakes, regulated environments.
I design and ship production agentic systems, and know what “reliable” actually means where mistakes have real consequences. A decade inside regulated financial institutions taught me what auditors, risk teams, and compliance actually require before software goes live. I now apply that to building AI systems that need to earn the same trust.
About
I build production agentic AI systems, with a focus on what makes them reliable enough to deploy where failure isn't an option. My path here is unusual: twelve years inside regulated financial institutions, followed by two years building and operating production AI systems independently. That combination is the lens I bring to AI reliability, evaluation, and governance: hands-on builder, with lived experience of what regulated, high-stakes environments demand. I've since applied this to production agentic systems spanning regulated financial workflows and revenue/lead-generation automation. This site shares my projects, technical writing, and experiments in AI, infrastructure, and reliable system design.
Focus Areas
AI Agents & Reliable Systems
Designing LLM-driven agents and tooling that automate research, analysis, and decision support. Each system is built with audit trails, scoped permissions, and human-in-the-loop checkpoints that high-stakes deployment requires.
Full-Stack Web Development
Building production web applications end to end with React, Next.js, TypeScript, and PostgreSQL.
Cloud, DevOps & Data Infrastructure
CI/CD pipelines, infrastructure as code, containerized deployments, and reliable data plumbing on AWS and Azure.
Regulated-Industry Domain Expertise
Twelve years inside cash, liquidity, and currency risk management at major banks and insurers. I know what regulated institutions actually require before they will trust new technology.
AI Safety, Security & Governance
Capable models are no longer the hard part. Trustworthy ones are. As AI moves from demos into treasury, banking, and other high-stakes workflows, the work that matters is closing the gap between what a model can do and what it does reliably, safely, and accountably. I write about how to evaluate agents, govern their behaviour, and report results honestly.
Navigating the Future of Agentic AI Evaluations: From Static Prompts to Dynamic Sandboxes
An end-to-end guide to evaluating agentic AI: the capability-reliability gap, the four pillars of evals, dynamic sandbox testing, and the governance reforms reshaping how benchmarks are reported.
Projects
Personal projects and experiments.
AI-powered financial interface for market insights, interactive charts, and contextual analysis.
AI-assisted follow-up and sales workflow that helps small B2B service businesses turn more inbound inquiries into booked calls.
Experiment in AI-assisted web search, research workflows, and summarization.
Web summarization tool that extracts and condenses long-form content.
Writing
Technical notes and write-ups, grouped by theme.
AI & Machine Learning
Cloud, DevOps & Security
How I Work
I take on remote contract and fractional work. That covers hands-on agentic builds, fractional advisory for teams navigating regulated environments, and evaluation or governance consulting where you need someone who has seen these systems from the inside. I work solo or as a technical lead embedded in a small team. Remote-first, based in Zurich, open to relocating for the right long-term role.
Contact
Currently available for remote contract and fractional work. If you are building AI systems that need to work reliably in a regulated or high-stakes environment, or evaluating where AI actually fits in your ops stack, I am happy to talk. A 15-minute call is usually enough to figure out if there is a fit.
Thanks for stopping by. 🚀