Data RAG Engineer

Meitner Energy; United States; On-site; USD 140,000 - 210,000; Full time

Job summary

Meitner Energy in the United States is seeking a Data and Retrieval Engineer to own the retrieval layer of its private AI platform. You will drive ingestion, parsing, chunking, and metadata for a large corpus of regulatory and engineering documents.

You will implement hybrid retrieval with dense vector search and sparse keywords, ensure access control before retrieval, and develop the evaluation harness to measure retrieval quality.

Qualifications

• Bachelor's degree in a technical field or equivalent experience
• 3+ years in data or ML engineering with focus on document processing and retrieval systems
• Strong Python and production data-pipeline experience
• Hands-on with production vector databases and embedding model deployment
• Experience with OCR tooling and parsing for complex document collections
• Familiarity with retrieval quality evaluation, hybrid retrieval, and reranking models
• Understanding that access enforcement precedes retrieval in regulated environments
• Ability to work on-site in Dallas.

Responsibilities

• Build and own end-to-end ingestion and processing pipeline for regulatory/engineering documents, including OCR, chunking, and metadata generation
• Implement hybrid retrieval combining dense vector search with sparse keyword search, and tune reranking for engineering/regulatory content
• Maintain a gold retrieval evaluation set and contributions to the platform's evaluation harness
• Ensure access-control labeling at document and chunk level and enforce access before retrieval

Skills

Python

Data pipelines

Vector databases

Embedding models

OCR tooling

Document parsing

Retrieval evaluation

Access control

Regulated environment

Education

Bachelor degree

Job description

Meitner Energy is seeking a Data and Retrieval Engineer to build and own the retrieval layer beneath the company’s private AI platform. This is a hands-on engineering role covering the full path from raw document to cited answer: ingestion and parsing, chunking and metadata, hybrid retrieval and reranking, access-control labeling, and the evaluation harness that measures whether any of it works. The position suits a data or machine learning engineer who treats retrieval quality as a measured engineering outcome rather than a configuration setting.

About Meitner Energy

Meitner Energy is developing advanced nuclear energy solutions intended to deliver reliable, scalable, and carbon-free electricity for industrial and grid applications.

Our international team brings together nuclear, engineering, commercial, regulatory, and project-development experience. We are building an organization focused on disciplined engineering, responsible execution, and the deployment of nuclear energy at meaningful scale.

The Opportunity

Meitner’s AI platform is only as useful as the retrieval layer beneath it, and this role owns that layer end to end. You will build the ingestion pipeline that processes a large corpus of regulatory and engineering documents, implement the hybrid retrieval and reranking stack that surfaces the right passage from the right document, and maintain the evaluation set that tells you whether retrieval is actually improving. The corpus includes scanned engineering drawings, technical standards, and regulatory material across multiple jurisdictions, so the parsing and optical character recognition problems are substantial and real.

Because the corpus contains export-controlled and otherwise sensitive material, every retrieval decision carries an access-control dimension. Access enforcement happens before retrieval rather than as a filter applied afterward, and citations must trace back to the source document and page. You will own the retrieval quality metric, the evaluation set behind it, and the pipeline that keeps both current as the corpus and the models change.

What You'll Do

Document Ingestion and Processing

Build and maintain the end-to-end ingestion pipeline for the regulatory and engineering document corpus, covering text extraction, optical character recognition of scanned material, chunking strategy, and structured metadata generation. You will handle document types that resist naive parsing, including engineering drawings, dense tables, and figures, developing extraction approaches that preserve the structural information retrieval depends on. Access-control labeling is applied and maintained at the document and chunk level during ingestion, so that classification metadata is accurate and auditable before any content reaches the index.

Retrieval and Reranking

Implement hybrid retrieval that combines dense vector search with sparse keyword search across a production search and vector retrieval stack, and tune the blend for the specific characteristics of an engineering and regulatory corpus where exact identifiers matter as much as semantic similarity. You will integrate and tune reranking models to improve precision on the results returned to the model, weigh the latency cost of doing so, enforce access filtering before retrieval rather than after, and implement citation generation that traces every answer to a source document, section, and page.

Evaluation and Quality

Build and maintain a gold retrieval evaluation set that represents the queries the platform will genuinely face, and own the metrics and the cadence on which they are run. You will implement extraction quality checks that detect parsing failures, character recognition errors, truncated chunks, and metadata gaps before they propagate into the index, and contribute retrieval quality data to the platform evaluation harness so that regressions introduced by model or pipeline changes are visible early.

What We're Looking For

Required Qualifications

• Bachelor’s degree in computer science, engineering, or a related technical discipline, or equivalent professional experience.
• 3+ years of professional experience in data engineering, machine learning engineering, or a closely related role, with a focus on document processing and retrieval systems.
• Strong Python, with demonstrated experience building and operating production data pipelines.
• Hands-on experience with production vector databases and with embedding model deployment.
• Practical experience with optical character recognition tooling and document-parsing pipelines for complex document collections, including familiarity with their failure modes.
• Working knowledge of retrieval quality evaluation methodology, hybrid retrieval, and reranking models.
• Understanding of why access enforcement must precede retrieval in a regulated environment, and the ability to implement document-level access control in a vector search context.
• Ability to work on-site in Dallas.

Preferred Qualifications

• Working proficiency in Spanish. Spanish is preferred because the corpus and the collaborating teams span English and Spanish material.
• Experience with engineering or regulatory document collections such as technical standards, drawings, and compliance filings.
• Familiarity with access-controlled retrieval in a regulated or security-sensitive environment.
• Exposure to prompt engineering and an understanding of how retrieval quality affects generated output.
• Background in nuclear energy, defense, aerospace, or another regulated industrial sector.

The Candidate We Are Seeking

The strongest candidate will be an engineer who has made a difficult corpus genuinely searchable and can prove it with numbers. This may be an excellent next step for a data engineer, search engineer, or machine learning engineer who wants full ownership of a retrieval layer rather than a slice of one. Candidates should be prepared to discuss pipelines they personally built, the specific ways their source documents defeated their first approach, and how they measured the improvement.

Please do not include confidential, proprietary, export-controlled, or otherwise restricted information belonging to a current or former employer.