1. Remote, with daily working hours overlapping US Eastern Time (Florida)
2. Working English required

You will be part of a delivery team for a contract intelligence engagement. The system reads large sets of

related legal documents (master agreements, amendments, addenda, exhibits, schedules) and answers questions

about what those documents say and which terms are in force at a given date.

You should be an AI engineer who combines production agentic engineering with genuine commercial contracts

knowledge. Both are hard requirements. A strong AI engineer without the legal background will not be accepted for

this role, because the work includes building the evaluation sets that determine whether the system's answers are

correct.

You will be responsible for:

• Designing and building agentic systems with frontier LLMs: agent loops, tool use, agent skills, and MCP servers.

• Building the document understanding pipeline: OCR, advanced image parsing, and table extraction from scanned and poorly structured PDFs.

• Decomposing complex legal questions into steps the system can execute and verify.

• Building evaluation sets from real contract questions with known answers, and using them to measure accuracy, citation quality, and the system's ability to return "not stated" instead of guessing.

• Working directly with legal and commercial stakeholders to translate their questions into system requirements.

Representative Questions the System Must Answer

You should be able to reason through questions of this type in a technical interview:

• The master agreement caps liability at 12 months of fees, with carve-outs for confidentiality and IP indemnity. A data breach occurs under a security exhibit added by a later amendment. Is the exposure capped?

• Amendment 2 restated a section in full. Amendment 4 modified one sentence of that section. What is in force today, and what is the citation?

• The SOW sets a 30 day notice period and the master agreement sets 90 days. The master agreement has no order of precedence clause. Which governs?

• Does this agreement contain a most favored nation clause? If it does not, how was that established?

• Which obligations survive termination, and for how long?

• A defined term is set in the master agreement and redefined in a later amendment. Which definition applies to the indemnity in the SOW?

Technical Requirements

• Strong Python, with production AI systems shipped and operated, not prototype or notebook work.

• Hands on experience with frontier LLM APIs and a production agent SDK.

• Agent architecture: agent loops, tool use, agent skills, and MCP.

• Document AI: OCR, advanced image parsing, and table parsing from scanned documents.

• Evaluation and testing of LLM systems, including hallucination control and abstention behavior.

• Senior level autonomy: able to own an ambiguous problem end to end without supervision.

Legal and Contracts Requirements

This is the requirement most candidate profiles will fail, and it is not waivable. You must have real working

knowledge of commercial contracts, including:

• Contract structure: master agreements, amendments, addenda, SOWs, exhibits, and how they relate to each

other.

• Amendment chains, restatements, and reasoning about which version of a clause is in force at a given date.

• Order of precedence and conflicts between documents.

• Common commercial clauses: liability caps and carve-outs, indemnity, termination, auto-renewal and

notice periods, assignment and change of control, survival, governing law, pricing and escalation terms.