Tilt Global set out to do something no one had done before: create an AI-powered coaching assistant that could detect decision-making biases without offering direct advice. Their mission was ambitious and technically untested.
With just two founders and limited funding, they needed to:
“Juan Carlos (Celerik CEO) understood our financial situation and offered to delaypayment—something we hadn’t expected. That flexible partnership set Celerik apart.”
— Brennon Martin, Co-Founder, TiltGlobal
We ingested real Decision-Making Lab (DML) transcripts into CosmosDB and built a Python-based API to tokenize text and run bias detection using GPT-4o. This provided fast, context-rich outputs while preserving data structure and performance integrity.
We crafted a clear, interactive dashboard in React/Vite that highlighted bias patterns and dynamics within teams. We ran design iterations every 48 hours, incorporating feedback from five early users to ensure the demo aligned perfectly with Tilt Global’s narrative.
While the initial POC remained local and didn’t implement cloud automation or CI/CD,it was structured with clear paths for scalability. Future phases include moving infrastructure to Azure and enabling real-time coaching interfaces.
“We accomplished the objective of proving our vision’s technical feasibility with minimal cost and effort. Now we know our AI coaching model is viable and ready to scale.”
— Brennon Martin, Co-Founder, TiltGlobal