AI Engineer

Farsight AI offers a hybrid work arrangement in New York, NY, with a competitive annual salary of $200,000 to $300,000. In this role, you will help build agentic AI systems and data pipelines that produce client-ready deliverables such as decks, models, and memos, while advancing evaluation, retrieval, and platform capabilities. You will transform proprietary platform data and client data into evidence-backed insights and contribute to robust grounding and verification across outputs.

Technologies you will work with include Python and frontier model APIs.

Benefits

• Salary range: $200,000 to $300,000 annually, commensurate with experience.
• Comprehensive medical, dental and vision insurance
• Unlimited PTO

Responsibilities

• Develop and own end-to-end agentic systems that generate client-ready financial deliverables — decks, models, and memos — from data intake to final output.
• Implement evaluation and quality gates for outputs, including deterministic document checks, model-led reviews, grounding and fidelity verification, and regression suites for non-deterministic results.
• Turn platform and client data into evidence-backed insights by building pipelines that discover, quantify, and verify patterns, feeding findings back into how drafts are created and checked.
• Build and refine retrieval systems that ground deliverables in both unstructured and structured financial data, ensuring every figure traces to a source.
• Advance agentic capabilities on the Farsight platform, focusing on composability, advanced function calling, and modular orchestration.
• Shape future work by prioritizing customer value and balancing time-to-market, execution risk, and impact.

Requirements

• At least 2 years deploying machine learning or large language model systems in production.
• Experience building agentic systems, including function calling, tool use, orchestration, and the reliability engineering that keeps multi-step workflows on track.
• Proven ability to design and validate evaluations for LLM systems; you know that asking a model to rate itself is not a sufficient answer.
• Experience building retrieval over unstructured and structured data and working with documents as structured artifacts (mixed text, numeric, and tabular content; understanding document formats and their internals; analyzing rendered outputs).
• A practical AI-assisted development workflow with a verification story proportional to delegated tasks.
• Strong Python proficiency and current fluency with frontier model APIs and their capabilities.