Applied AI Engineer

United StatesRemotemid

OtherApplied Ai Engineer

2 views0 saves0 applied

Apply Now

Quick Summary

Key Responsibilities

OpenHands, Langchain or coding agent frameworks,

Technical Tools

OtherApplied Ai Engineer

Jito's engineering team has already rebuilt how it works around AI, and the gain in output is real. Everywhere else in the company, that same gain sits uncaptured. This role exists to go get it, with a company-wide mandate from day one.

Jito builds the market layer of Solana. We run the validator client behind a majority of the network's stake, JitoSOL, BAM, and JTX. About 40 people. Ops, legal, finance, marketing, business development, and product each have well-understood, expensive workflow problems that a strong engineer could solve in a week, and today those problems sit unowned or get picked up as side work.

You would be the first dedicated AI engineer at Jito Labs.

We're a small team with product-market fit. That means:

High ownership: every team member drives significant technical decisions and shapes our roadmap.
Fast execution: we ship rapidly and iterate based on real-world feedback. Progress, not motion.
High quality bar: the details are the product. We don't trade them for speed.

You'll work with the CEO and with teams across the company to find the highest-leverage work. Some of that is problems people already know they have. Much of it is optimization they've stopped noticing, in processes that function well enough that nobody has questioned them. This role owns the cross-cutting investments no single team will prioritize, and builds for the teams that can't build for themselves.

The default is to buy and configure where something off the shelf does the job, and to build where nothing does. Everything you ship carries a named maintainer and clear criteria for when to sunset it.

A meaningful part of the role is spent with the people doing the work, engineers included, in enough depth to understand where their time actually goes. From there you scope it, build it, secure it, roll it out, and keep it running.

Security owns the platform layer. You own the applications built on top of it, in partnership with Security from design time forward.

This is a mid-to-senior individual contributor role, roughly three to five years of experience. It is not a research or model-training position.

Reports to the CEO.

The surface area is the whole company. Every team has a different workflow, a different data shape, and a different tolerance for a tool that gets it right most of the time. Doing this well means understanding how Jito actually runs, team by team, including the optimizations that people inside a process have stopped seeing.

Buying is usually correct and rarely sufficient on its own. Finding the seam between configuring a vendor and building it yourself is most of the judgment this role requires.

Adoption is a people problem before it's an engineering one. The strongest model users at any company are well ahead of the median, and closing that distance takes teaching and documentation as much as code.

Security and speed pull against each other, constantly. We are infrastructure that validators and stakers depend on, and we take that seriously, so anything touching internal data gets designed with Security from the start. Judging where the extra week is worth spending, and where it isn't, is a large part of this role.

Three to five years shipping and operating production services. You can stand up VM sandboxes, MCP servers, and vendor LLM integrations, and you'll pass an engineering screen.

LLM and agent tooling you've shipped that people use today. We'll ask who uses it and how often. Prototypes nobody adopted won't clear the bar.

Buy-first instincts with the range to build. You can explain your last custom build and why it was necessary, and you've sunset projects of your own.

The ability to sit with someone in Legal, Finance, or Marketing, extract the real workflow, and ship something they use every week, without a PM translating for you.

Judgment about which teams are actually drowning, and the restraint to avoid pushing tools on people who didn't ask.

Security at design time by reflex, with constraints treated as parameters of the design.

You think like an operator. The most interesting problem and the highest-impact problem are frequently different ones, and you can tell which is which. Getting nerd sniped by a hard technical problem that saves the company an hour a month is the failure mode here.

Self-directed prioritization. You surface use cases, estimate their impact, rank them with the CEO, and say no. There is real work waiting on day one, and the expectation is that you find more of it than anyone hands you.

Operational ownership. You maintain what you ship, and have clear kill criteria for each project to avoid unnecessary maintenance burden.

The prospect of 40 people producing the output of 400 reads as interesting to you.

Blockchain experience is not required.

If you have experience in any of the following, you may be a good fit:

OpenHands, Langchain or coding agent frameworks, deployed and maintained in a real environment

Building and operating MCP servers

CLI tooling and automation that other engineers picked up and kept using

Security and network operations: identity, secrets management, egress control, sandboxing

Connecting existing systems into something that holds up under real use, and recognizing when to write it yourself instead

Being the only person in your function at a company under 50 people

Owning a vendor surface: spend, data exposure, renewals

Rust, Python, TypeScript or a systems background