AI Front-End Developer, Senior

AI Front-End Developer, Senior

The Opportunity:

G RACE is ARPA-H's production AI assistant. We've shipped the first generation of agentic AI, and now we're moving quickly to go deeper: user-built agents, multi-agent orchestration, and a platform teams can extend themselves on. We are a small, startup-minded team that ships fast and owns what we build end-to-end. We are looking for a senior front-end-leaning full-stack developer to own the experience layer of G RACE: the chat interface, the rich rendering of AI output, and the interaction patterns that make an agentic system feel trustworthy and fast.

The front end is where users decide whether G RACE is reliable and worth using. You will own that surface, while operating within the privacy, security, and responsible-use requirements of our mission and the federal environment.

What You'll Work On:

• Own the G RACE front end end-to-end, including the Chainlit-based chat application and the React or Type Script components and interaction patterns layered on top of it.
• Own the rendering pipeline for rich AI output, including streaming responses, markdown, code blocks, tables, citations, diagrams with Mermaid, maps with GeoJSON, and generated images, so that complex output is legible and trustworthy.
• Design and build interaction patterns for agentic workflows, including tool-call progress, multi-step reasoning states, human-in-the-loop confirmations, and graceful handling of long-running or partial responses.
• Apply industry best practices and emerging front-end patterns specific to agentic AI, including streaming UX, agent and tool-call visibility, and interfaces for user-built and multi-agent workflows.
• Own client-side state, streaming, and real-time updates, including Web Socket or SSE, so the interface stays responsive while the backend is still working.
• Build file upload, document preview, and results-display experiences for G RACE's knowledge search and research workflows.
• Work close to the application layer in Python where Chainlit lives and customize rendering, session handling, and interaction behavior beyond what the framework gives you out of the box.
• Partner with backend engineers on the API contract the front end consumes, shape it so it serves the user experience, and implement the glue when the boundary is fuzzy.
• Implement front-end-facing integrations with G RACE's data sources and APIs, including LLM provider outputs, knowledge search, and tool results.
• Write readable, tested, reviewed, and documented production-quality code across the stack.
• Work hand-in-hand with our human-centered researcher and designer to turn user insights and design intent into interfaces that actually serve the user, not just approximations of the mockup.
• Translate research findings and usability feedback into concrete UI changes and close the loop by shipping, measuring, and iterating.
• Treat design fidelity, interaction detail, and edge-case behavior as part of the job and sweat the small things because they are what users feel.
• Own front-end performance from the user's perspective, including time-to-first-token, perceived latency, and smooth streaming under real network conditions
• Ensure accessibility, including Section 508 and WCAG, is built in, not bolted on.
• Build front-end observability, including client-side error tracking, usage analytics, and the instrumentation needed to know what users actually do.
• Rapidly prototype and iterate on new G RACE capabilities and ship, measure, and improve.
• Establish and improve front-end coding standards, design review, and testing practices.
• Communicate technical decisions clearly to both engineers and non-engineers.
• Mentor and unblock other engineers with a bias toward ownership and speed.
• Ensure strong privacy, security, and compliance in all UI, data handling, and integrations.

You Have:

• 7+ years of experience in software engineering, building and operating production web applications, including on the front end
• 5+ years of experience designing and building for AI, LLM, chat, or agentic products, including streaming responses, citation and source display, tool-call visibility, and building for uncertain or probabilistic output
• 5+ years of…