Keep pulling the thread on Angelo Stracontanio.
The Apprentice 4.1 model was created by fine-tuning a state-of-the-art foundation model using reinforcement learning and 12 years of Apprentice's domain-specific manufacturing data.
AT&T implemented an AI agent for network management that reduced the time required to triage a network problem to 2% of the previous time, a 50x improvement.
To ensure reliability and consistency in a manufacturing setting, Apprentice constrains its AI agents to operate within tightly defined workflows with a narrow set of possible outcomes.
Apprentice's manufacturing execution system (MES) is used to produce two-thirds of the world's commercially approved gene therapies.
Apprentice's AI tools can reduce the time spent on quality review by a third or more, which directly reduces cost of goods sold and improves time-to-market for customers in specialized industries.
Apprentice developed its own proprietary AI model, named Apprentice 4.1, for its manufacturing solutions.
Apprentice found that general off-the-shelf AI models were inadequate for manufacturing applications because they lacked specificity, consistency, and compliance.
Apprentice's A1 platform is composed of multiple specialized sub-agents designed for specific roles within a manufacturing facility, such as operator, quality, process engineering, and site leadership.
Apprentice's initial market focus was on the life sciences industry, where product quality is critical due to the risk of patient harm.
Apprentice developed a proprietary evaluation system for its AI models that specifically measures for consistency, a critical metric for manufacturing applications.
Apprentice's platform includes optional 'guardrails' that require a human operator to approve or reject actions proposed by the AI agent before they are executed.
To support human-in-the-loop oversight, Apprentice's AI provides a full source history detailing the data and reasoning that led to a proposed action.