4 months ago
Base Salary
$160k - $200k/yr
Responsibilities
- Design and implement production systems integrating LLMs into real product workflows.
- Build prompt pipelines, tool integrations, structured outputs, and systems combining models with traditional software components.
- Develop reliable LLM architectures with safeguards, retries, evaluation loops, and observability.
- Create evaluation datasets and tooling to measure model performance and prevent regressions.
- Design and optimize retrieval pipelines using chunking, indexing, vector search, and knowledge-retrieval improvements.
- Collaborate with product, design, and engineering teams to ship polished AI-powered product features.
- Contribute to system architecture, code reviews, testing, and engineering practices that maintain platform reliability and scalability.
Requirements
- At least five years of experience building production systems.
- Strong software engineering fundamentals and experience designing reliable, maintainable, and scalable systems.
- Production experience building systems around large language models, including prompt design, context management, tool orchestration, and improving output reliability.
- Experience building evaluation frameworks, datasets, and experiments for prompts, models, and system architectures.
- Experience with retrieval-augmented generation, embeddings, vector databases, and semantic search.
- Conceptual understanding of transformers, attention mechanisms, and modern language-model architectures.
- Understanding of LLM strengths and limitations, including hallucinations, prompt sensitivity, context limits, latency, and cost.
- Ability to write clean, maintainable, well-tested code and contribute to architecture and code reviews.
Benefits
- Meaningful equity opportunity for early team members.
- Flexible choice of hardware and development setup, including high-performance Mac laptops.
- Flexible paid time off.
Categories
About Ravenna
Helping IT and Operations teams deliver incredible support to their colleagues since 2024.