5 months ago
Remote, WorldwideSenior
Responsibilities
- Design and implement AI agent features and extend existing agents with new capabilities.
- Manage agent context using sub-agents, retrieval-based context management, sliding context windows, and related techniques.
- Integrate and evaluate models from OpenAI, Anthropic, and Google, including assessing new model releases and capabilities.
- Enable agents to safely call external tools and APIs for web searches, database queries, information retrieval, and domain-specific actions.
- Build multi-step AI workflows using frameworks such as Vercel AI SDK and LangGraph.
- Collaborate with engineers and product managers to deliver production-ready, efficient, maintainable, and monitored AI features.
- Collect and curate datasets from agent responses and multi-turn conversations, and analyze behavior, failure modes, and success signals.
- Experiment with prompting, context handling, fine-tuning, and other techniques to improve AI output quality and reliability.
- Develop systematic tests and automated LLM evaluation benchmarks using user feedback and performance data.
- Share knowledge and support mid-level and junior engineers.
Requirements
- Hands-on experience working with LLMs and building LLM-powered applications or systems.
- Familiarity with TypeScript and willingness to work in TypeScript daily.
- Solid software engineering fundamentals and experience building production-ready systems.
- Deep understanding of prompt engineering practices across different tasks and models.
- Ability to work autonomously, own tasks end-to-end, and manage complex work with minimal oversight.
- Strong analytical, problem-solving, data collection, and data analysis skills for improving agent behavior.
- Strong verbal and written English communication skills.
- Familiarity with DSPy is a bonus.
- Machine learning fundamentals and experience with model evaluation metrics are a bonus.
- Open-source AI/ML contribution or maintenance experience is a bonus.
- Experience reading and implementing AI/ML research techniques is a bonus.
Benefits
- Fully remote work arrangement
- Globally distributed team
- No college degree required
- Applicants do not need to be located in the United States
Tech Stack
Categories
About Bolt.new
Bolt.new builds an AI-powered, browser-based platform to create, edit, and deploy full-stack web and mobile apps via chat and agent workflows. Offered by StackBlitz, it targets developers and product teams who want instant cloud development environments and push-button deployment without local setup. The privately held company is headquartered in San Francisco and was founded in 2017.