Software Engineer, AI/ML

Interfere•New York, NY

22h•Onsite

About The Position

Interfere is building a product-quality platform that detects user-facing product failures in real time and gives teams the context they need to understand and fix them. The company is a seven-person team in New York, has raised $5.1 million, and is already working with design partners in production. The Software Engineer, AI/ML will own the intelligence layer behind the product, building agents, evaluations, and inference pipelines that identify what is broken, determine causes, and help teams decide what to do. The work combines research and production engineering, with systems running against real codebases and runtime data. The role includes developing agents and their sandboxes for investigating codebases, traces, logs, and runtime behavior; LLM pipelines for triage, root-cause analysis, and fix proposals; model routing; anomaly detection; evaluation, dataset, observability, and feedback systems; and retrieval, indexing, context construction, and integrations that help models understand customers’ codebases. The engineer will also build systems that learn from bugs found and fixed, improving future detection. The team works in person in New York City, ships daily, and values deliberate engineering, ownership, and measurable results. Work is measured by outcomes rather than desk time, though the company notes that the urgency and interest of the work often lead people to put in serious hours.

Requirements

Has shipped ML- or LLM-powered systems into production where real users depend on the output.
Can move comfortably between research and engineering and learn unfamiliar technologies quickly.
Can prototype with new frameworks or techniques and use evaluations to determine whether the result is better.
Can own a project from 0 to 1, define useful next steps for ambiguous AI problems, and ship without a fully specified plan.
Treats evaluations as a first-class engineering problem and uses measurement rather than intuition for systems making decisions for users.
Has production experience with agent architectures, tool use, multi-step reasoning, or autonomous workflows.

Nice To Haves

Background in code understanding, program analysis, or systems that reason over source code.
Experience building retrieval, RAG, or context-construction systems for large or messy data.
Experience with anomaly detection, time-series modeling, or learned monitoring systems.
A deep understanding of the inner workings of AI/ML systems.
An agentic system built for use by other engineers or operated in production at meaningful scale.
Open-source work, technical writing, or research demonstrating technical depth.
Experience shipping agents or LLM features into production and informed views on what works and what remains unresolved.
Experience designing and running evaluations that materially changed a product or research direction.
Experience starting a research direction, product feature, or internal tool from zero.
Evidence of quickly learning an unfamiliar domain and shipping real work in it.

Responsibilities

Build agents that reason over codebases, traces, logs, and runtime behavior to find and explain bugs before users encounter them.
Build the sandboxes, browsers, and web-research capabilities used by agents to investigate.
Develop LLM pipelines for triage, root-cause analysis, and automated fix proposals, including model routing.
Build systems that detect anomalies in product behavior, performance, and user experience across noisy real-world data.
Create evaluations, datasets, observability, and feedback loops to measure and improve system performance.
Architect standards for the agentic parts of a product that is still being defined.
Build retrieval, indexing, context construction, and integrations that help models understand customers’ codebases.
Develop systems that use bugs found and fixed as signals to improve future detection.
Own ambiguous AI problems and take work from definition through implementation and delivery.