AI Engineer — Document Intelligence

FinStack; Hyderabad; On-site; INR 1,200,000 - 2,100,000; Full time

Job summary

FinStack is developing applied AI for document-heavy banking operations, including OCR, vision-language models, extraction, and retrieval-augmented generation. You will build and evaluate document classification, OCR and extraction pipelines, and integrate LLMs for structured extraction and reasoning.

The role requires 3+ years building ML/AI systems in production, strong Python, and hands-on OCR/document layout or vision-language model experience.

Qualifications

• 3+ years building ML or AI systems in production or close to production.
• Strong Python; experience with modern LLM and document AI tooling.
• Hands-on experience with OCR, document layout or vision-language models.
• A rigorous approach to evaluation and an honest view of model limitations.

Responsibilities

• Work on applied AI for document-heavy banking operations: OCR, vision language models, extraction, consistency checking and retrieval-augmented generation.
• Build and evaluate document classification, OCR and extraction pipelines.
• Integrate and prompt vision language models and LLMs for structured extraction and reasoning.
• Design evaluation datasets and metrics; measure results against expert review.
• Build RAG-based assistants with citations, access control and logging.
• Package and serve models for on-premise and private-cloud deployment.

Skills

Python

LLMs

OCR

Document AI

Evaluation metrics

RAG-based assistants

Job description

You will work on applied AI for document-heavy banking operations: OCR, vision language models, extraction, consistency checking and retrieval-augmented generation. The emphasis is on measurable accuracy, evidence for every output and deployments that run in private environments.

Build and evaluate document classification, OCR and extraction pipelines.

Integrate and prompt vision language models and LLMs for structured extraction and reasoning.

Design evaluation datasets and metrics; measure results against expert review.

Build RAG-based assistants with citations, access control and logging.

Package and serve models for on-premise and private-cloud deployment.

3+ years building ML or AI systems in production or close to production.

Strong Python; experience with modern LLM and document AI tooling.

Hands-on experience with OCR, document layout or vision-language models.

A rigorous approach to evaluation and an honest view of model limitations.