What Does an AI Engineer Do Now? - Aparna Dhinakaran, Arize AI

AI Engineer · 15 min · 126 sentences · from YouTube's caption track

Each timecode opens YouTube at the start of that sentence. Line anchors (#s42) are the cue ids in the WebVTT, and every line carries its start and end seconds. All transcripts has every talk, and the whole corpus as one file.

00:01[music]
00:12Awesome.
00:13Well, welcome to AI engineer.
00:15Um uh thought given where at AI engineers New York, we'd kind of revisit the question of what does an AI engineer do?
00:24Feel like every year we have an idea of what an AI engineer does and that definition kind of changes.
00:31Um And as I was putting this talk together, it was super timely because Karpathy actually just tweeted something this weekend where he kind of he kind of talks about this growing
00:41gap between basically folks at the frontier who are running thousands of agents and swarms of always-on agents and he really kind of drops this funnel where 6 billion plus users have maybe touched or interfaced
00:57with an LLM.
00:581 to 2 billion or maybe like casual free tier users.
01:0220 million are professionals who have maybe started using it for more taking some piece of work that would have taken weeks down to to days.
01:11But really those at the frontier labs are which is a super small percentage.
01:16It's like 5,000 users.
01:19Uh are the ones who are running thousands and thousands of agents at once.
01:25And as I look at this, it really is I mean he's kind of dropped a bit of the funnel.
01:30I mean the first question anyone looks at this is that where am I on this?
01:33And you know, he's kind of drawn out a bit of a road map of how does everyone get to in the next couple of years one of those 5,000
01:42users who are running thousands of agents.
01:45And this has been happening I'd say for for a couple of decades now where the abstraction layer has gotten higher and higher.
01:52I mean definitely faster in the last couple of years, but we've gone from engineers writing higher level abstraction languages to I mean just copy-pasting code, using local agents to using
02:07cloud agents to really the expectation now is how how do all of us start running a team of agents or factories of agents internally and build toward that.
02:18[snorts]
02:18And what this has led to is essentially what is a then and now of what an AI engineer is.
02:24And this is is this is a bit uh I I think of a of a roadmap for some, but I think internally this is what we've already started to kind of push towards.
02:34As an AI engineer, your job's no longer to just write the feature, write an AI feature, it's about how do you build the loop to write that feature.
02:42It's no longer about you know, just reading the diff and actually doing code review, it's actually how do you set up systems and agents to do the code reviews, and how do you write evals to actually do
02:53the non-deterministic um to review those non-deterministic tasks.
02:58It's actually no longer even about just setting up metrics and and software kind of telemetry, but it's actually how do you have agents go through and read the observability
03:07data and have agents actually consume that telemetry to find patterns.
03:13And so,
03:13[snorts]
03:14this loop that we've kind of been talking about, which is the build, the eval, ship, you monitor, you improve it, well, the AI engineer is now actually responsible not just for doing each of those steps,
03:25but how do you build the loop that does the that that actually does each of those steps.
03:31And at Arize, one of the things we've rolled out in our product, and I'm going to talk about a bit of what we've learned in building this, is we built signal
03:38to actually run this loop for AI engineers.
03:41What signal does is that signal actually looks at all of the traces, it reads the telemetry, it identifies patterns in that telemetry, and it can actually put up a fix or a PR
03:53if it has access to the repo.
03:56And signal signal's an always-on agent.
03:59So, this is an example of how signal starts, but it actually starts off by just reading the traces.
04:06Super simple example, but it goes through, reads um and all of this configurable, but and certain volume of traces that came in in some cadence.
04:16It reads it and then the goal is to actually identify issues from that telemetry data.
04:22It has to keep track of issues that it's previously seen.
04:25So, we have to consider the memory of did see something before that it thought was an issue?
04:32Did something seem like an issue, but it wasn't actually captured and now as it collects more data, it actually thinks it's much more of a critical critical issue.
04:40And so, signal actually starts off by reading the traces to identify problems.
04:46And so, one of the first things that we realized was that the future of who's actually consuming telemetry data has really changed.
04:55It's no longer about humans consuming telemetry data.
05:00It's not even about humans consuming it and agents kind of taking on the task, but it really is how do you set up your telemetry data so that agents are the primary consumers of that data.
05:11And what that does is that that observability data can become the trigger for actually running and triggering these always-on agents.
05:19And so, the first thing that if you're building a team of of you know, a factory of really always-on agents is what triggers these agents to wake up or start.
05:29And there can be a variety of triggers.
05:31For signal, for us it's you know, it's configurable to run on an ongoing cadence.
05:36So, every 3 hours, 6 hours, go pull the last, you know, X amount of traces, read them, identify problems.
05:45It can also be triggered by events.
05:48So, Slack messages, um if some user complains about something, it can be a metric-based trigger, but these are essentially triggers to kick up the agent actually starting.
06:01Once the agent actually starts, we have to spend some time figuring out, well, what is what is an agent run actually look like?
06:07For us, we actually pull in all of the context of the traces, um any additional third-party data, so we allow collecting to APM logs, APM metrics, um but all of the third-party telemetry data that we want to pull in, um
06:23access to the repo, um access to the actual um the the access to any keys that are needed to actually run run this agent, but all of this information
06:37is um put inside of a sandbox, and this is actually what's started up every time the agent wakes up.
06:44And so, the agent's triggered, we build a sandbox, we put everything that we need inside of that sandbox, there's a proxy, the only way to get information in and out of the sandbox is by this single proxy,
06:55and everything that happens inside of that sandbox is actually already getting traced, and every output, so the issues that are generated and created are all getting saved to some kind of memory.
07:07And so, this happens every single time the agent actually runs.
07:11Every time the agent finishes running, so it swept through all of the traces, um it either found something or it didn't, there's what actually stays and what actually gets thrown away.
07:24What stays is that if it found any issues, all of the output is actually saved off in some kind of memory, so that the next time an agent gets spun up, it doesn't have to start from ground zero.
07:34It has a history of what are the issues that it actually found.
07:38Um the agents are traced, so all of the telemetry data of what the agent called, um you know, if it actually put up a PR, all of those things are actually traced.
07:49And any skills that it may have learned while actually doing a run can also get saved to the actual repo.
07:56But, the things that go away is the sandbox gets torn down, any credentials, um you know, obviously get get removed, um the repo access is removed, and so all of the stuff that doesn't need to be saved kind of gets torn down after the run of signal.
08:13And we've designed signal in a way to actually keep this agent configure durable.
08:18So, that means that either if the if the sandbox fails, or if the harness fails, then it shouldn't have to start everything from scratch in the middle of the run.
08:29It should actually just be able to pick up from the step that it failed, and all of the state, so-called, is kind of saved off in kind of a a memory outside of the sandbox.
08:41And so, we've designed this to actually be a very durable agent, so that if it fails, it can spin it back up.
08:48Um you can launch a new sandbox, launch a new harness to kind of pick back up where it started.
08:54So, all of this is essentially the underpinnings of how we built signal to run every single run.
09:00Agent gets triggered, agent gets started, loaded with all the contexts, the evidence, uh it sweeps through the telemetry data, it either finds something or it doesn't, all of that gets saved to memory.
09:10If something fails in the middle, the agent can get reset up.
09:14And we built signal to also be you know, agnostic of the model or the harness.
09:19So, today signal supports not just Cloud Code, which is what we started with, but it also supports Cursor and Codex.
09:26And the goal is to actually support all of the the major, even the open source harnesses out there.
09:31Um and signal actually also supports um multiple uh it will support kind of multiple models.
09:39So, even if you're using Claude code, uh you can pick any of the models inside the Anthropic family.
09:45Um and this is really important to us because one, it's an awesome way for us to evaluate the quality of the agent um every time we spin it up.
09:55It helps us control costs, so we can see, you know, as we use more powerful models, the better quality maybe the PRs are, but can we reduce the quality of the the you know, can we reduce the cost and actually still maintain the rigor
10:10um and maintain the quality of issues that are actually detected by Signal.
10:15So, this is something we deeply thought about from from first principles when we were designing Signal.
10:21And this is actually a pretty good segue into what does an AI engineer need to do?
10:26Well, an AI engineer really needs to kind of learn how to write evals.
10:30And so, we actually every time Signal runs, Signal finds an issue, it puts up a PR.
10:35Well, there's evals that are run on top of the agent to test that PR.
10:40Um and an eval can be as simple as um you know, testing something as part of your development workflow.
10:47You have a dataset, you can be a simple prompt um it with with outputs that you're testing to make sure that there's no regressions.
10:55Um but these evals can also they need to be kind of aligned with human judgment.
11:02And I'd argue that this is actually probably where an AI engineer should be spending a lot of their time as they're building the system to actually run the loop.
11:11They should be actually spending time looking at the traces with their agents, identifying what the common errors are.
11:18Um and there's a lot of skills I think I was hearing Lance yesterday talk about the Claude skill to build evals.
11:23We have a skill with the arise um evaluator to actually help you build and write evals as well as align those evals with the human judgment.
11:33But, this is probably where the shift to defining what is good becomes really, really important for an AI engineer to actually do.
11:42And for a lot of teams I I think we speak with and we work with, their question becomes, "Well, how do I trust the evaluator?"
11:49And do I do I really trust that the eval is um you know, is going to capture you know, how how do I trust that it's as good as my judgment?
12:01And I'd say what we spend a lot of our own time doing for our own products is that we actually evaluate the evaluators.
12:08Um we spend a decent amount of time on actually building these data sets, golden truth data sets.
12:16We run e-vals, verify them against the human output, and then try to align to make sure that the judge actually agrees with the human um and and they're in agreement before we actually ship out
12:30uh a new evaluator.
12:31And these e-vals can then be used both in an offline testing as well as in an online testing manner.
12:38Um and this this is probably what I would say as an AI engineer is more and more of what the job is becoming.
12:47Um and so, what I'm going to leave you guys with is as an AI engineer, you've built this agent, you've built this agent to actually run, be always on,
12:56you run e-vals before you actually ship it into production.
13:01You can ship it, and then once it goes into production, the goal is to actually trace everything.
13:07And tracing everything now means not just tracing the agents, but also tracing the coding agents behind the agents that built this.
13:16So, you can trace um cloud code, you can trace cursor, you can trace codex, but all these coding agents can be traced so you can understand what happened um in the factory of agents that you all are building.
13:31And then these agents that all the traces can then also be evaluated online.
13:35So, there's an offline evaluation when you're actually testing what you're what a new change is, but as it goes into production, these become your online evals that you're running on top of the telemetry data.
13:48This for us ends up becoming the new data that then gets pulled back into the agents to read the telemetry data, the evals, and surface up kind of problems
14:01inside of the traces.
14:02So, the the system really becomes how do you go from collect data, eval that data, run agents on top of it so that you can find problems and figure out where the gaps are to actually improve
14:16improve your systems.
14:18So, if I go back to kind of the question that we started this whole thing with, it's what does an AI engineer do?
14:24I mean, the gap between I think there's a pretty big gap between what's happening in the labs today and what's happening with a lot of businesses.
14:32And some of it could just be the data inside of businesses are a lot more messy than what is very constrained and testable inside of a lab.
14:40But, AI engineers are kind of the the role to close that gap, and I would think it really starts with tracing everything, tracing the agents you're actually building, running evals, so being able to learn how to test these
14:54test these products that we're really building, and then how do you build a system so that agents can actually improve from the telemetry that's actually being collected so that you're no longer just building individual features,
15:07but you're actually building the system to improve to improve your products.
15:13If you want to learn more, come check out we have a booth.
15:15You can check out Signal, learn more, but thank you so much for listening to us.
15:20[applause]
15:20[music]
15:26Hey.