current sales pipeline is fully human-based between lead acquisition and the closing call. I'm leading a team applying data-science and statistical methods to find best candidates and execute automation via llm pipelines. we also defined appropriate benchmarks aligned with the business logic to properly measure AI impact.
defining agent-first pipelines to obtain processes descriptions from directors. humans are prompted by AI agents and their responses feed our OKF adherent wiki. this process depends on a MCP server to transport proper skills and my requirements of fulfilled knowledge.
via provided knowledge to our wiki, i'm leading a team to cross-check information and find biggest gaps and consequences to revenue (either lost or not-realized gains). we defined a llm-based pipeline to estimate lowest-hanging fruits in order to guide and substantiate our directors work.
dspy-based all in one RAG-powered enterprise chat app. sunsetted after serving hundreds of users (b2c+b2b) across about five brazilian SMBs
news parser, feed aggregator and fully functioning agentic news room and AI powered clipping
complete AI-powered video conferencing. video calls + state of the art agentic orchestration to provide useful summaries specific to your workflow. sunsetted after over 100k hours served to over 200 users
science-backed algorithm to determine energy levels during the day to real users. based any wearables data (oura, whoop, apple watch and others). included analysis on several sleep metrics and a custom llm-based insights generation to real consumers. at eon2life
developed a machine learning model to determine if the patient should go to the emergency room or to a video consultation. end product reached 97% accuracy and is currently deployed in production. this took place at unimed-bh in a partnership with indigo hive