ANII | INTELLIGENT ASSISTANT FOR FORMS AND AI TECHNICAL TRAINING

THE COMPANY

The National Agency for Research and Innovation (ANII) of Uruguay is the public organization responsible for funding and promoting scientific research and technological innovation in the country. It manages calls and application programs that connect the academic, business, and government sectors, handling a significant volume of technical forms and associated documentation in each funding cycle.

THE CHALLENGE

ANII’s application forms are extensive and technically demanding documents. Errors in completion, incorrect file uploads, and the need to manually draft publishable summaries created an operational bottleneck for both applicants and internal teams. At the same time, the organization identified the need to develop in-house AI capabilities to operate and evolve solutions without external dependency.

THE SOLUTION

Baufest developed and implemented two complementary solutions. First, an intelligent assistant integrated into the application process: automatic field validation with configurable rules (date ranges, list lookups, text similarity), real-time verification of attachments, automatic generation of the Publishable Summary using generative AI, and a chatbot with a knowledge base trained on ANII’s official documentation. The architecture follows a Document Intelligent Process approach with specialized agents and dynamic rule management without requiring deployments. Second, Baufest designed and delivered two technical training programs for the systems team: an initial series of 14 sessions covering fundamentals and development with LLMs (LangChain, RAG, LLMOps, agents with LangGraph and CrewAI), and a second series of 6 advanced sessions focused on production architectures, including advanced RAG with HyDE and hybrid search, structured outputs with Pydantic, agentic systems with self-correction and planning, testing with LLM-as-judge, CI/CD, observability with LangSmith, and compliance in government environments.

BENEFITS

• Cleaner application process: applicants detect and correct errors before submission, reducing rework for internal teams.
• Automatic generation of the Publishable Summary using AI, eliminating a high-effort manual task in each call.
• Reduced workload for the support team thanks to a chatbot with a knowledge base based on official documentation.
• Real-time verification of attachments, preventing rejections due to incorrect or non-compliant documentation.
• Internal technical team trained through 20 sessions on LLMs, agents, RAG, LLMOps, and AI application production.
• Technological autonomy: ANII now has the in-house capabilities to operate, maintain, and evolve AI solutions without vendor dependency.