Exequiel Silvestre

About

AI/LLM engineer with production experience delivering large-scale RAG and voice-agent solutions for banking clients. Implemented a SharePoint-based, non-technical content update workflow with incremental hourly ingestion into a vector DB, and actively contributes to Microsoft’s open-source GPT-RAG accelerator while using modern orchestration (Semantic Kernel, LangGraph) and LLM observability/evaluation tooling.

Key Strengths

• Built and deployed production RAG solutions for banking clients
• Designed SharePoint-driven self-service content updates with hourly incremental re-ingestion
• Pragmatic tradeoff decisions for voice agents (accuracy vs real-time latency)
• Latency reduction via audio payload optimization and prompt-level latency analysis
• Improved perceived latency with conversational UX techniques (status/intermediate messages)
• Hands-on LLM orchestration experience (GPT-RAG, Semantic Kernel, LangGraph)
• Uses LLM observability and evaluation practices (LangSmith/Langfuse, LLM critic, user testing, latency/token cost tracking)
• Strong client-facing discovery and translation of business needs into AI architecture (bank voice agent)

Experience

Education

Skills

Languages