This includes writing skill client libraries, their documentation, and testing them end to end. - Build and operate sandboxed execution environments where agent-generated code runs with tenant-level security boundaries and durable file persistence. - Ship full-stack features across frontend, backend, skill libraries, and infrastructure.
Requirements
The AI Core team builds the platform that gives users AI-driven access to Chainalysis's blockchain investigation and compliance products.
We build the agent runtime, the execution infrastructure, and the tool and skill systems that let AI reason about what users need and deliver answers, artifacts, and automation grounded in Chainalysis's proprietary data across sandboxed environments, durable workflows, and a deep catalog of product integrations.
We run a small team with high ownership; we're looking for engineers who care about building systems where AI has real tools, real data, and real consequences.
Our approach is different from most enterprise companies utilizing AI.
That means designing how agents discover and use capabilities across Chainalysis's product suite, building the sandboxed environments where agent-generated code executes with tenant isolation and scoped credentials, writing the skill libraries that give agents structured access to blockchain investigation and compliance data, and shipping the interfaces where users interact with results.
You own the full lifecycle from design through production operation. - Develop the workflow platform that lets users automate complex processes beyond single conversations. - Drive AI quality, cost, and observability, including multi-model orchestration, context window management, prompt engineering, and tracing across multi-step agent runs. - Raise the reliability and security bar for systems where AI acts on behalf of users.
You've hit the failure modes and learned from them. - A working understanding of how to structure what an agent knows and can do within the constraints of a context window.
- Deep backend engineering fundamentals in a typed language: concurrency, error handling, state management, and system design.
- Delivered frontend experiences with a modern stack, component architecture, data fetching, and real-time streaming.
You might also have: - Designed tool interfaces or capability systems for AI agents, skills, function schemas, or similar patterns for giving models structured access to external systems - Worked with the Effect ecosystem or strongly-typed functional programming in TypeScript - Built multi-tenant platforms or systems with enterprise security requirements -
Experience in Python library development or designing SDKs for programmatic consumption - Worked within Blockchain, compliance, or financial services domains Technologies we use: - TypeScript, Effect, Node.js - React, TanStack Router, TanStack Query, Tailwind, shadcn - Various AI Frameworks (Vercel AI SDK, Agent SDK, OpenCode) - Python - PostgreSQL - AWS Lambda, S3, STS, CloudWatch, ECR - OpenTelemetry - Docker, pnpm, Turborepo, Vitest - Amazon Bedrock, Google Vertex AI About Chainalysis Blockchain
Businesses and governments around the world are using blockchains to make banking more efficient, connect with their customers, and investigate criminal cases.
As adoption of blockchain technology grows, more and more organizations seek access to all this ecosystem has to offer.
We provide complete knowledge of what’s happening on blockchains through our data, services, and solutions.
With Chainalysis, organizations can navigate blockchains safely and with confidence. You belong here.
Contact
You can learn more here https://go.chainalysis.com/rs/503-FAP-074/images/Interview%20Accommodations%20Request.pdf.
Additional details
Instead of sprinkling LLMs into every feature, we give the model scoped access to Chainalysis's data and product capabilities and let it write and execute code on behalf of the user in sandboxes with tenant isolation, scoped credentials, and full auditability.
The user describes what they need; the system figures out the procedure.
That turns every customer and every person in the organization into a problem solver, able to surface insights, build investigation graphs, generate reports, and automate workflows that previously required both domain expertise and programming ability.
We're building a system that can extend itself, where new capabilities don't always require new code from us, because the agent can compose what already exists to solve problems no one anticipated.
You'll shape how agents solve problems based on user intent, and you'll own what you ship from design to production operation.
We’re looking for candidates who have: - Built and shipped full-stack features across frontend, backend, and infrastructure, with demonstrated ownership from design through production operation. - Worked with LLMs in production and formed real opinions from it.
Tool use, agent orchestration, streaming, prompt engineering, or retrieval-augmented generation.
You've thought about the tradeoffs between baking knowledge into prompts, exposing tools, and letting agents write code against documented APIs.
- Built or operated systems with real security constraints, tenant isolation, scoped credentials, sandboxed execution, or similar patterns where trust boundaries matter.
You've been on-call, debugged production incidents, and improved the reliability of systems you own.