Applied AI Research Engineer

Starburst

10h•$215,000 - $270,000•Hybrid

About The Position

You will own the intelligence layer that makes AIDA's agents correct, trustworthy, and measurably better over time. The work spans information retrieval, knowledge representation, and evaluation science. You will turn ambiguous notions of "agent quality" into clear metrics, build the grounding systems that connect agent reasoning to verified data, and create the evaluation infrastructure that makes quality a first-class engineering discipline. You will operate at the research/systems boundary: running experiments with academic rigor and shipping results with production engineering discipline. Research and engineering are not separate tracks here. You will own experiments end to end, from hypothesis through production deployment. As an Applied AI Research Engineer at Starburst, you will: Design and build grounding systems that connect agent reasoning to verified enterprise data sources Build and optimize retrieval pipelines (RAG, hybrid search, structured query generation) for accuracy and latency Define data representation strategies that preserve semantic fidelity across heterogeneous enterprise data (catalogs, schemas, lineage) Create evaluation frameworks: automated benchmarks, regression suites, human evaluation protocols Convert validated research findings into production systems that ship to users Establish quality metrics and dashboards that track agent correctness week over week Build feedback loops where user interaction data flows back into evaluation datasets and informs grounding improvements

Requirements

3+ years of experience in information retrieval, NLP, knowledge representation, or applied ML research
Production experience building RAG, grounding, or retrieval systems (not prototypes or demos)
Strong evaluation methodology: benchmark design, statistical analysis, reproducible experiments
Comfort operating at the research/systems boundary: you read papers and you ship code
Python fluency; experience with vector databases, embedding models, LLM APIs
Track record of converting research insights into shipped production systems

Nice To Haves

Experience with enterprise data systems (SQL engines, data catalogs, schema metadata)
Familiarity with text-to-SQL or structured query generation
Published research or open-source contributions in IR, NLP, or evaluation methodology
Experience designing evaluation pipelines that run in CI/CD
Familiarity with JVM-based systems

Responsibilities

Design and build grounding systems that connect agent reasoning to verified enterprise data sources
Build and optimize retrieval pipelines (RAG, hybrid search, structured query generation) for accuracy and latency
Define data representation strategies that preserve semantic fidelity across heterogeneous enterprise data (catalogs, schemas, lineage)
Create evaluation frameworks: automated benchmarks, regression suites, human evaluation protocols
Convert validated research findings into production systems that ship to users
Establish quality metrics and dashboards that track agent correctness week over week
Build feedback loops where user interaction data flows back into evaluation datasets and informs grounding improvements