S
Sr. AI Engineer
Surround Sound Media
1 hour ago
Full-time
On-site
New York City, New York, United States
Build the agent systems behind a working local-news network.
SurroundSound builds and operates AI-powered local news outlets covering local government, schools, business, and sports. Our agents research and draft stories, and human producers review and publish them.
A prominent Silicon Valley VC backs us, and we're
already generating revenue.
You'll work on context and retrieval, tool orchestration, durable state, source provenance, and evaluations. Work directly with the CTO to shape the architecture, own delivery, and learn from the producers using what you ship. The goal is useful local coverage with less repetitive work for the newsroom.
What you'll do
Design agent workflows:
retain reporting context across runs and coordinate research, drafting, and review. Evaluations:
turn producer corrections and failure traces into evaluation cases and tested changes to prompts, retrieval, and tools. Ship:
handle duplicate work, interrupted runs and cost.
What you'll bring
Experience shipping an LLM or agent application used by real users.
Internal tools, small products, and startup work count. Be ready to explain your contribution, how people used it, and what you improved after launch. Strong backend engineering fundamentals: Python, SQL, APIs, and background jobs. You can debug and maintain production software. Practical experience with model/tool integration, structured outputs, retrieval, context and workflow state. Sound judgment about AI output: checking evidence, recognizing unsupported claims and evaluating changes before release. Clear communication, independent problem-solving, and effective use of AI coding tools. You can explain, review, and validate the code you ship.
Nice to have
Media, publishing, or other products where information accuracy matters. Search, entity resolution, document extraction, or long-running workflows. AWS, FastAPI, Celery, PostgreSQL, or enough TypeScript/React to contribute to producer tools.
To apply:
Share your résumé and include a brief example of an AI application you've shipped: who used it, what you owned, and a failure you fixed. Links to working products or code are welcome.
A prominent Silicon Valley VC backs us, and we're
already generating revenue.
You'll work on context and retrieval, tool orchestration, durable state, source provenance, and evaluations. Work directly with the CTO to shape the architecture, own delivery, and learn from the producers using what you ship. The goal is useful local coverage with less repetitive work for the newsroom.
What you'll do
Design agent workflows:
retain reporting context across runs and coordinate research, drafting, and review. Evaluations:
turn producer corrections and failure traces into evaluation cases and tested changes to prompts, retrieval, and tools. Ship:
handle duplicate work, interrupted runs and cost.
What you'll bring
Experience shipping an LLM or agent application used by real users.
Internal tools, small products, and startup work count. Be ready to explain your contribution, how people used it, and what you improved after launch. Strong backend engineering fundamentals: Python, SQL, APIs, and background jobs. You can debug and maintain production software. Practical experience with model/tool integration, structured outputs, retrieval, context and workflow state. Sound judgment about AI output: checking evidence, recognizing unsupported claims and evaluating changes before release. Clear communication, independent problem-solving, and effective use of AI coding tools. You can explain, review, and validate the code you ship.
Nice to have
Media, publishing, or other products where information accuracy matters. Search, entity resolution, document extraction, or long-running workflows. AWS, FastAPI, Celery, PostgreSQL, or enough TypeScript/React to contribute to producer tools.
To apply:
Share your résumé and include a brief example of an AI application you've shipped: who used it, what you owned, and a failure you fixed. Links to working products or code are welcome.