You will be responsible for developing two critical components that ensure the integrity of our finance agent. First, you will create a retrieval system that enables the agent to accurately answer questions based on Voyage’s finance policy documents, incorporating strict access controls and proper citation. Second, you will design a set of deterministic, non-AI validators that identify and prevent incorrect financial data from reaching human approval.

Key Responsibilities:
- Develop a document ingestion and hybrid retrieval pipeline for Voyage’s finance policy documents.
- Implement access control mechanisms prior to ranking, ensuring users without permission cannot see restricted documents or citations.
- Create citation and provenance tracking to guarantee every AI-generated response is linked to a specific, access-verified source.
- Build deterministic financial validators, including balanced-entry checks, fiscal-period validation, and match-tolerance logic, using exact decimal arithmetic to maintain precision with financial data.
- Design an abstention guardrail that prompts the agent to decline answering when evidence is incomplete or contradictory, avoiding guesswork.
- Maintain and update the evaluation dataset used for regression testing the agent’s performance.

Required Qualifications:
- Minimum of 3 years’ experience in information retrieval, retrieval-augmented generation (RAG) systems, or natural language processing (NLP) pipelines.
- Strong attention to numeric accuracy, with comfort working in exact decimal arithmetic for financial data rather than floating-point calculations.
- Ability to collaborate closely with non-engineer finance domain experts to define and handle edge cases effectively.

Preferred Qualifications:
- Experience building retrieval-augmented generation systems that incorporate document-level access control.
- Background in finance, accounting, or ERP-related domains is highly desirable.

This position requires working on-site to facilitate close collaboration and effective communication within the team.

Job Details

Total Positions:
1 Post
Job Shift:
First Shift (Day)
Job Type:
Job Location:
Gender:
No Preference
Age:
18 - 65 Years
Career Level:
Mid-Level
Maximum Experience:
2 Years
Apply Before:
Oct 01, 2026
Posting Date:
Sep 25, 2026

Modulus OS

· 11-50 employees - Lahore

What is your Competitive Advantage?

Get quick competitive analysis and professional insights about yourself
Talk to our expert team of counsellors to improve your CV!
Try Rozee Premium
I found a job on Rozee!