AI Engineer

Nuvena; New Delhi; On-site; INR 900,000 - 1,800,000; Full time

Job summary

Nuvena in New Delhi is building Navio, the Execution Operating System, to extract obligations, owners, dates and penalties from contracts. You will teach Navio to read contracts, design, test and improve the AI that drafts clauses and answers questions across documents.

You own the AI from the first prompt to the feature customers rely on, building test sets, tracking metrics and fixing issues. You will collaborate with engineers to deploy reliable Python services and keep outputs safe and

Qualifications

• 2-4 years of software engineering with hands-on work on large language models.
• Strong Python skills and real experience with LLM APIs and prompt design.
• Experience building evaluation for AI features and improving results.
• Understanding retrieval with embeddings and vector search.

Responsibilities

• Build and improve the extraction pipelines turning contracts into structured data (obligations, owners, dates, proof, penalties).
• Write, version and refine prompts for extraction, classification, review and drafting.
• Develop evaluation sets with the legal team and measure accuracy after changes.
• Design retrieval across large document sets: chunking, embeddings, search and citations.
• Collaborate with backend engineers to turn AI workflows into reliable Python services with clear APIs, logging and monitoring.
• Keep outputs safe, consistent and easy for a user to check, while controlling cost and response time.

Skills

Python

LLM APIs

Prompt design

Structured outputs

Software engineering

Evaluation development

Embeddings

Vector search

Ownership

Writing clearly

Job description

About Nuvena: Nuvena builds Navio, the Execution Operating System: every promise a company makes or is owed — in a contract or to a team — gets an owner, a follow-up and proof.

Teach Navio to read contracts: design, test and improve the AI that finds obligations and drafts clauses.

New Delhi - 2-4 years

Role Summary

You teach Navio to read contracts. You design, test and improve the AI in Navio OS that pulls obligations, owners, dates and penalties out of real agreements, drafts clauses and answers questions across documents. You treat accuracy as something you measure, not something you hope for: you build the test sets, track the numbers and fix what breaks. You own the AI from the first prompt to the feature customers rely on.

Key Responsibilities

• Build and improve the extraction pipelines that turn contracts into structured obligations, owners, dates, proof requirements and penalties.
• Write, version and refine prompts for extraction, classification, review and drafting, and know why each one works.
• Build evaluation sets with the legal team, measure accuracy on every change, and catch regressions before they ship.
• Design retrieval across large document sets: chunking, embeddings, search and citations, so every answer points back to the clause it came from.
• Work with the backend engineers to turn AI workflows into reliable Python services with clear APIs, logging and monitoring.
• Keep outputs safe, consistent and easy for a user to check, while keeping cost and response time under control.

Must Have

• 2-4 years of software engineering, including hands-on work building on large language models.
• Strong Python, and real experience with LLM APIs, prompt design and structured outputs.
• Has built an evaluation for an AI feature and improved results because of it.
• Understands retrieval: embeddings, vector search and how chunking changes the answers.
• Reads carefully and writes clearly; spots the clause the model missed.
• Takes ownership and finishes what you start.

Good To Have

• An interest in law, contracts or business documents.
• OCR and parsing for scanned PDFs, tables and messy documents.
• Fine-tuning, or running open-source models in production.
• Shipped AI features to enterprise customers.

What we expect from everyone:

• Go-getter
• Problem-solver
• Open-minded
• Takes decisions
• Owns the outcome
• Works as one team