AI Engineer

Sentrion, Inc.; New York, Northern (NY, KY); Hybrid; USD 120,000 - 170,000; Full time

Job summary

Sentrion, Inc. in New York is building the software behind our account research platform. You’ll develop services and data workflows to collect, process and retrieve company information, handle incomplete records, and ensure safe retry of failures.

You’ll design and validate AI features for extracting and interpreting information, manage context sent to models, and build tests and prompts. Collaboration, reliability and performance improvements will be part of the role.

Qualifications

• Experience shipping and maintaining software used by others.
• Strong programming fundamentals and backend systems knowledge (APIs, databases, asynchronous tasks, testing).
• Experience with language models, structured outputs, and context limits.

Responsibilities

• Build services and data workflows that collect, process and retrieve company information.
• Develop AI features for extracting and interpreting information; validate outputs and manage model context.
• Create evaluation cases to test prompts, models and processing logic.
• Improve reliability, response times and processing costs with data-driven fixes.
• Write maintainable code and participate in design, testing and releases.
• Turn customer feedback into reproducible examples and fix data, code or model behavior.

Job description

We’re building Sentrion to help teams find companies that need what they offer, and understand the problem before reaching out. A business might be opening a new office, rebuilding a department or changing the way it operates. Our work is to make sense of those changes in the context of what each customer sells. We want the team reaching out to know how it can help, and the company hearing from them to have a reason to listen.

About the role

You’ll build the software behind Sentrion’s account research. A good result should help someone understand a company and decide whether they can help it. That gives you a practical question to work from when improving a feature: what would the customer be able to do better if we got this right?

The work includes the services that process company information, the model calls that interpret it and the features that make the result understandable. You’ll need to look closely at the outputs as well as the code. When a result misses the point, work out whether the source was incomplete, the application lost context or the model made an unsupported inference. Then decide how to fix it and verify the change.

What you’ll do

• Build services and data workflows that collect, process and retrieve company information. Handle incomplete records, changing source formats and failures that need to be retried safely.
• Develop AI features for extracting and interpreting information. Validate structured outputs, manage the context sent to a model and retain the evidence needed to check a result.
• Build evaluation cases around the questions customers need answered. Include cases where the available information doesn’t support a conclusion, and use these examples to check changes to prompts, models and processing logic.
• Improve reliability, response times and processing costs. Use logs and measurements to find the cause of a problem, then check that the change actually improves it.
• Take features through technical design, implementation, testing and release. Write code another engineer can maintain, review changes with the team and follow through on issues after launch.
• Turn customer feedback into a reproducible example before changing the system. Establish what a useful answer would have been, then fix the relevant data, code or model behavior.

What we’re looking for

• Experience shipping and maintaining software used by other people. You can talk through something you built, the decisions you made and what you learned after it was released.
• Strong programming fundamentals and comfort working on backend systems: APIs, databases, asynchronous tasks, automated tests and debugging across service boundaries.
• Hands-on experience building with language models beyond an isolated prompt. You understand structured outputs, context limits and the need to evaluate results against representative examples.
• The ability to inspect and reason about data. You can write queries, trace where a record came from and distinguish a data problem from an application or model problem.
• Good engineering judgment. You can explain when a straightforward implementation is enough, when more infrastructure is justified and how you would verify either choice.
• Clear communication about progress and problems. You ask for missing context, flag uncertainty and take responsibility for following a fix through.

Useful, but not required

• Search or retrieval systems, information extraction, or processing unstructured text.
• Background jobs and data pipelines, including retries, duplicate records and partial failures.
• Operating model-based features in production, with monitoring and evaluation alongside the application code.

Working at Sentrion

When you see a problem, spend some time working out a possible solution before bringing it to the team. You don’t need every answer, but you should have a view on what to try and why. If you disagree with a decision, including the founder’s, bring the facts and explain the approach you think would work better.

When your own decision turns out to be wrong, be clear about what happened and what you’ll do to put it right. Follow through, then look at what you could have done differently. That kind of self-reflection matters to us. This role is based in New York with a hybrid setup.