Thiago Mendes Garcia
Applied AI specialist building AI agents, conversational wor
New to Contra
Thiago is building their profile!
As a licensed Brazilian attorney and Applied AI specialist, I designed and experimented with Legal AI assistants and workflow structures focused on improving legal research, document analysis, information organization, and professional decision support.
My work included designing specialized prompt architectures, defining legal reasoning instructions, structuring knowledge and context, creating response standards, and testing AI behavior across different legal scenarios.
I also explored how AI can support document-intensive workflows by identifying relevant information, organizing complex content, improving research efficiency, and creating more structured legal analysis while maintaining human professional oversight.
This project combines my legal practice with hands-on AI experience, allowing me to approach Legal AI from both perspectives: understanding the technology and understanding the real workflows, risks, and requirements of legal professionals.
0
7
I developed an AI knowledge base and workflow automation structure designed to improve how AI systems access information, follow business logic, and support more reliable task execution.
My work included organizing knowledge into a clear structure, mapping information flows, defining process logic, and designing automation paths that could support conversational agents and business workflows more effectively.
I focused on making information easier for AI systems to retrieve, interpret, and use consistently, while also reducing ambiguity and improving workflow efficiency.
The objective of this project was to create a practical foundation for smarter AI operations, combining structured knowledge with better workflow design to support more accurate, scalable, and business-oriented automation.
0
11
I designed a conversational AI and prompt architecture focused on improving how LLM-powered agents understand instructions, maintain context, handle exceptions, and respond consistently across different user scenarios.
My work included structuring system prompts, defining instruction hierarchy, designing conversation flows, establishing behavioral rules, managing context, and creating fallback and exception-handling logic.
I also tested different conversation scenarios to identify ambiguity, conflicting instructions, context loss, and inconsistent responses. Based on these tests, I continuously refined prompts and agent behavior to improve reliability, clarity, and task completion.
The objective was to create a structured conversational framework that goes beyond basic prompting and provides a more predictable, context-aware, and business-oriented AI experience.
0
14
I created KinkLoth.AI (http://KinkLoth.AI) to design practical AI agents for real business operations, especially customer service and conversational workflows.
For this project, I personally structured the agent’s prompt architecture, conversation flows, business rules, knowledge base, context handling, fallback logic, exception scenarios, and response evaluation process.
The goal was to move beyond a simple chatbot and build an AI agent capable of understanding context, following business rules, handling unexpected situations, and delivering more consistent customer interactions.
I also tested the agent across different scenarios, identified behavioral failures, refined prompts and instructions, and continuously improved the workflow based on real conversational challenges.
1
19