Switchboard AI

I founded Switchboard AI to solve a fundamental inefficiency in how organisations use AI: the challenge of selecting the optimal model for each specific task. Switchboard is an intelligent routing service that automatically directs user prompts to the most suitable AI model from providers like OpenAI, Anthropic, Google, and others. By matching requests to models based on factors such as task complexity, cost constraints, latency requirements, and domain expertise, Switchboard enables organisations to achieve better results at lower costs whilst eliminating the cognitive overhead of manual model selection. The platform employs a sophisticated routing architecture built on transformer-based neural networks and Bayesian optimisation, trained on extensive performance data from LMArena and real-world usage patterns. The system operates on a freemium model designed for mid-market businesses, serving development teams requiring efficient API orchestration and business users across finance, consulting, law, and HR who need reliable AI assistance without technical complexity. The infrastructure emphasises low-latency inference and continuous learning from routing decisions to improve accuracy over time. I conceived Switchboard after observing the proliferation of AI models and the resulting decision paralysis organisations face when implementing AI solutions. Having completed my Master's in Computing at Imperial College London with a focus on machine learning, I recognised this as both a technical challenge and a significant business opportunity. I spent nine months developing the platform and successfully onboarded initial B2B clients, though achieving full product-market fit remains an ongoing pursuit. The experience has deepened my expertise in AI orchestration, enterprise software, and go-to-market strategy within the rapidly evolving AI landscape.