Benjamin Cowen contrasts frontier APIs with custom training, describing tradeoffs between quick iteration and control over cost, latency and business-specific performance.
He identifies rising API costs and stalled evaluations as possible reasons to consider fine-tuningFine-tuning adapts a pretrained model by updating its parameters on examples chosen for a particular task or subject., while cautioning that teams need suitable data and mature evaluations before trainingEvaluation is the systematic assessment of a system's behaviour or results against explicit criteria..
He discusses open-source training libraries, serverless compute, parallel reinforcement-learning rollouts and model serving. His examples and product claims reflect his experience at Modal.
Watch on YouTube
