AI Engineer

Life Science Logistics•Dallas, TX

7d•Onsite

About The Position

We are seeking an AI Engineer to establish and maintain a structured prompt library for common use cases like summarization, Q&A, extraction, code generation, and file analysis. You will apply advanced prompting techniques, design and build RAG pipelines connecting various company systems, and deploy and tune LLM-powered applications. This role involves continuous A/B testing of prompt variants, documenting performance benchmarks, and serving as the subject-matter expert on Claude Desktop's file-handling capabilities. You will create reusable prompt patterns for multi-file inputs and long-context documents, champion prompt engineering best practices across internal teams, and embed guardrails for PII handling, data residency, and AI safety. Additionally, you will test for prompt injection risks specific to local file inputs and build internal evaluation harnesses to measure prompt quality and consistency.

Requirements

3+ years of software engineering or applied AI experience, with at least 1 year focused on LLM prompt engineering.
Deep hands-on experience with Claude, GPT-4, or similar large language models in production or near-production settings.
Background in technical writing, UX wireframing, instructional design, or developer advocacy.
Prompt engineering discipline including system prompt design, zero-shot, few-shot, output validation, hallucination mitigation.
Experience with adversarial prompt testing, red teaming methodologies, or AI safety evaluation frameworks.
Production deployment mindset. You monitor what you build, you own uptime, you care about latency and cost.
Excellent written communication skills; ability to translate complex technical concepts for non-technical audiences.
Proven ability to collaborate cross-functionally across engineering, product, and customer-facing teams.
Direct experience with Claude Desktop, including MCP (Model Context Protocol) server configuration and local file tooling.
Familiarity with RAG (Retrieval-Augmented Generation) architecture and vector databases (e.g., Pinecone, Weaviate, pgvector).
Proficiency in Python or JavaScript for building prompt pipelines, evaluation scripts, and automation tooling.
Supply chain, logistics, or 3PL domain knowledge including WMS, EDI 850/810/856, DSCSA familiarity.
Knowledge with Intelligent Document Processing (IDP), OCR pipelines, and handwritten text extraction (AWS Textract, CargoShot, or equivalent).
Must be able to successfully pass all preliminary employment requirements (i.e., background check and drug screen).

Nice To Haves

Previous experience working with a lean team in a regulated industry, such as pharma, healthcare, government, or financial services is preferred.
Exposure to government or enterprise RFP processes, understanding of compliance documentation, and proposal requirements are nice to have.

Responsibilities

Establish and maintain a structured prompt library for the company which will cover common use cases (summarization, Q&A, extraction, code generation, file analysis).
Apply advanced prompting techniques including chain-of-thought, few-shot examples, role specification, and XML-structured inputs.
Design and build RAG pipelines connecting our WMS, EDI logs, SOP repositories, contract data, and other systems.
Deploy and tune LLM-powered applications including internal knowledge assistants, client-facing chat, extend RAG based response repositories, and leverage AI to optimize workflows, processes, and drive system improvements.
Continuously A/B test prompt variants and document performance benchmarks.
Serve as the subject-matter expert on Claude Desktop's file-handling capabilities, including referencing local PDFs, Word documents, spreadsheets, and code files within prompts.
Create reusable prompt patterns that work reliably with multi-file inputs, long-context documents, and structured data.
Champion prompt engineering best practices across internal teams including Operations, Control Tower, Business segments, and General Counsel to translate business problems into AI solutions.
Embed PII handling rules, data residency constraints, jailbreak resistance, and refusal behavior guardrails into production prompt workflows.
Test for prompt injection risks specific to local file inputs — including malicious content embedded in PDFs, DOCX, or CSV files uploaded through Claude Desktop.
Build and maintain internal evaluation harnesses to measure prompt quality, consistency, and regression over model updates.
Other duties as assigned.