Cognitive Systems: Receipt OCR, RAG Knowledge

The client engaged Taller in 2023 for engineering capacity on a multi-year platform modernization. Taller ran multiple pods through 2026, from investment-management delivery to Azure platform engineering, plus embedded Scrum Masters, TPMs, and practice leadership.

The client was building out an ambitious AI portfolio (including a receipt-scanning claims product, an AI knowledgebase, and a personalization engine), and every part of it depended on secure, well-built identity services underneath: the systems that verified who a member was and what they were allowed to do. Building those services in a regulated financial environment was demanding and slow. The code had to meet strict security and compliance standards, and the client also had to expand its engineering capacity to keep pace with the wider AI roadmap without lowering that bar or slowing delivery.

Taller embedded engineers in the client’s Identity team to design, build, and maintain these identity services. To move faster without cutting corners, the team used GenAI Copilot tooling (AI that assisted engineers as they wrote code) to generate and accelerate the C# and Node.js work behind authentication and authorization. Taller brought relevant depth to this kind of work from its wider portfolio: a set of production generative-AI engagements for a major US airline, spanning advanced search and retrieval, a system for comparing versions of technical manuals, and a framework for multi-agent AI workflows, a strong reference point for building AI responsibly in regulated, document-heavy environments.