AI Engineer (LLM)
ΠΡΡΡ & Π‘ΠΎΠΏΡΠΎΠ²ΠΎΠ΄
ΠΠ»Ρ ΠΌΡΡΡΠ° Ρ ΡΡΠΎΠΉ Π²Π°ΠΊΠ°Π½ΡΠΈΠ΅ΠΉ Π½ΡΠΆΠ΅Π½ Plus
ΠΠΏΠΈΡΠ°Π½ΠΈΠ΅ Π²Π°ΠΊΠ°Π½ΡΠΈΠΈ
TL;DR
AI Engineer (LLM): Building and optimizing retrieval pipelines, agent workflows, and model evaluation systems with an accent on RAG and agent orchestration. Focus on implementing hybrid retrieval, ensuring tenant isolation, and developing robust evaluation datasets for production LLM systems.
Location: On-site in Atlanta, Georgia
Company
is a global staffing and services firm providing professional talent and technical solutions.
What you will do
- Contribute to hybrid retrieval pipelines, implementing and tuning components and validating results on real evaluations.
- Build and maintain agent orchestration layers, including tool integrations and prompt management.
- Instrument AI and agent systems for observability by tracing model calls and capturing telemetry.
- Create evaluation datasets and test suites to establish CI-level quality gates for retrieval and agent workflows.
- Support document AI capabilities, focusing on parsing, extraction, and OCR pipelines.
- Implement per-tenant isolation checks to ensure no cross-tenant data leakage occurs.
Requirements
- 3+ years of software engineering experience.
- 1+ year of experience building production LLM-powered systems (RAG pipelines, agent workflows, or fine-tuning).
- Proficiency in Python, including async code, data pipelines, and REST APIs.
- Practical experience with LLM evaluation methodologies and observability tooling.
- Working knowledge of modern LLM and information-retrieval concepts.
- Must be based in or able to work on-site in Atlanta, Georgia
Nice to have
- Experience with knowledge graphs or property-graph query languages.
- Exposure to document AI, specifically PDF parsing and table extraction.
- Open-source contributions to projects like LangGraph or sentence-transformers.
- Familiarity with prompt-injection risks and mitigation strategies.
ΠΡΠ΄ΡΡΠ΅ ΠΎΡΡΠΎΡΠΎΠΆΠ½Ρ: Π΅ΡΠ»ΠΈ ΡΠ°Π±ΠΎΡΠΎΠ΄Π°ΡΠ΅Π»Ρ ΠΏΡΠΎΡΠΈΡ Π²ΠΎΠΉΡΠΈ Π² ΠΈΡ ΡΠΈΡΡΠ΅ΠΌΡ, ΠΈΡΠΏΠΎΠ»ΡΠ·ΡΡ iCloud/Google, ΠΏΡΠΈΡΠ»Π°ΡΡ ΠΊΠΎΠ΄/ΠΏΠ°ΡΠΎΠ»Ρ, Π·Π°ΠΏΡΡΡΠΈΡΡ ΠΊΠΎΠ΄/ΠΠ, Π½Π΅ Π΄Π΅Π»Π°ΠΉΡΠ΅ ΡΡΠΎΠ³ΠΎ - ΡΡΠΎ ΠΌΠΎΡΠ΅Π½Π½ΠΈΠΊΠΈ. ΠΠ±ΡΠ·Π°ΡΠ΅Π»ΡΠ½ΠΎ ΠΆΠΌΠΈΡΠ΅ "ΠΠΎΠΆΠ°Π»ΠΎΠ²Π°ΡΡΡΡ" ΠΈΠ»ΠΈ ΠΏΠΈΡΠΈΡΠ΅ Π² ΠΏΠΎΠ΄Π΄Π΅ΡΠΆΠΊΡ. ΠΠΎΠ΄ΡΠΎΠ±Π½Π΅Π΅ Π² Π³Π°ΠΉΠ΄Π΅ β