Skip to content
aitrainer.work - AI Training Jobs Platform
Engineering Full-Time
NeuBird

Staff AI Engineer, Agent Platform (US)

NeuBird • Redwood City, California, United States

Company

NeuBird

Annual salary

$210k – $280k/yr

Location

Redwood City, California, United States

Listed

34d ago

Experience:
5 to 15 years
Workplace:
Fully on-site, 5 days a week, at the Redwood City, CA office.
Equity:
Negotiable. The amount isn't set yet, but the HM said it will be well compensated because this is a critical, high-impact role.

On-site in Redwood City, California, United States.

Send us your LinkedIn and CV. If your experience fits, we'll introduce you to the recruiter filling this role.

Apply for a referral →

The recruiter emails you before anything happens and may suggest other jobs that suit you better.

About this role

You'll own the agent platform: the harness that runs every investigation, the evaluation system that decides what ships, and the architecture that sets what the agent can do next. The role requires the combination of ML systems expertise and agentic AI expertise; either one alone isn't enough. It's a senior IC role with real design authority. You set technical direction for the platform and raise the engineering bar for the AI engineers around you.

What you'll do

  • The next architecture. Design how the agent builds knowledge across investigations and scales its reasoning across data sources, with the tenant isolation and grounding discipline enterprise data demands.
  • Harness reliability. Investigations run under real deadlines against dozens of integrations. You own checkpointing and recovery, graceful degradation, tool-use robustness, deadline and cost budgets, and the telemetry that shows how the agent itself fails.
  • Evidence-grounded output. Make sure the agent's confidence is earned by retrieved evidence, and that its recommendations fit each customer's actual tool stack.
  • Evaluation as the release gate. Design and evolve the system that tells us whether the agent got better: eval sets built from real incidents, regression testing, and quality metrics tied to every release.
  • Technical direction. Lead design reviews, set platform standards, and mentor AI engineers through code and design.

Nice to have

  • Distributed-systems depth: checkpointing and recovery, deadline design, cost and latency budgeting
  • Observability, AIOps, or SRE-tooling exposure: logs, metrics, traces, incident workflows
  • Operating agents under strict grounding or hallucination-control requirements in enterprise or regulated settings
  • Multi-tenant isolation of customer data in ML or agent systems
  • Retrieval or knowledge systems built over operational data

About NeuBird

When production breaks, our agent is already investigating. It correlates logs, metrics, traces, and change events across a customer's stack and hands the on-call engineer a root cause they can act on. Enterprise SRE and operations teams run it on live incidents today. Generally available since December 2024, it has resolved more than 1 million alerts, with 94% RCA accuracy and root cause in under five minutes.

Our founders, Gou Rao (CEO) and Vinod Jayaraman (CTO), previously built Portworx, the cloud-native storage company acquired by Pure Storage, their third successful exit. We're backed by Mayfield, Microsoft's M12, Xora Innovation, StepStone Group, and Prosperity7 Ventures. Read more on our blog.

Must-have background

We're looking for the combination of ML systems depth and agentic AI depth. Strong candidates bring both; we don't hire for one and hope to grow the other.

  • End-to-end ML/LLM systems depth in production: data, model or agent, deployment, and evaluation. You've owned systems where quality was measured, not assumed.

  • Strong production engineering fundamentals. You write the code that runs, debug it live, and own its reliability.

  • Evaluation-driven development as daily practice. You've designed eval sets, gated releases on them, and can name a metric you moved.

  • Hands-on depth in agent orchestration and/or persistent context or memory: multi-step tool use, parallel or delegated agents, cross-session state, retrieval-backed context. Expert in one, working depth in the other.

Why this role

You inherit a production agent with real enterprise customers and real incident data. You get a green-field mandate on the platform's next architecture, and a quality bar enforced by an eval gate rather than opinion. The platform decisions here are still one person's to drive. That person is you.

Common questions

How do I apply for the Staff AI Engineer, Agent Platform (US) role at NeuBird? +

Use the Apply for a referral button on this page to send us your LinkedIn and CV. If your experience fits, we'll introduce you to the recruiter filling this role. They'll email you to check you're interested, then put you forward for this job or others that suit you better. It's free.

What does this NeuBird role pay? +

The listing gives $210k – $280k/yr.

Is this role remote? +

The listing gives the location as Redwood City, California, United States. Fully on-site, 5 days a week, at the Redwood City, CA office.

AI Engineer interview prep

Questions AI Engineer interviewers often ask. They aren't taken from this listing.

How do you decide which evaluation metric to optimize for a classification model when the classes are imbalanced?

Accuracy is misleading on imbalanced data because a model can score well by always predicting the majority class. I look at precision, recall, and F1 for the minority class specifically, and pick the metric that matches the real cost of false positives versus false negatives in the deployment context, since those costs are rarely symmetric in practice.

What is your process for debugging a deep learning model that trains fine but performs poorly at inference time?

I start by checking for a train/inference mismatch: different preprocessing, batch normalization behaving differently in eval mode, or data leakage during training that inflated validation scores. Comparing the exact input pipeline used at training time against the one used at inference usually surfaces the discrepancy before I need to touch the model architecture itself.

How do you approach feature selection when working with high-dimensional tabular data?

I start with correlation and mutual information analysis to drop obviously redundant features, then use a tree-based model's feature importance as a first pass filter. From there, recursive feature elimination with cross-validation tells me where the marginal value of additional features drops off, which keeps the final feature set both smaller and more robust to overfitting.

Explain how you would detect and address data drift in a production machine learning pipeline.

I monitor the statistical distribution of input features and model outputs over time, using something like population stability index or KL divergence against a reference window. When drift crosses a threshold, I retrain on recent data rather than the full historical set, and I keep an alert in place so drift is caught before it shows up as a quality regression.

See all 10 AI Engineer interview questions →

Related roles

NeuBird

Browse all referral roles

Salaried roles at startups, AI companies and established businesses, all filled by referral. One application covers every role.

See all referral roles →