Fortune 500 Tech Company Enhances AI Capabilities with Hugo

A Fortune 500 tech company known for its content-sharing mobile apps needed to enhance its foundation models for North American users.

Challenge

The client faced difficulties in:

• Improving caption accuracy for platform content.

• Refining prompt-response interactions for virtual agents.

• Navigating language nuances and cultural references specific to North American audiences.

Solution

Hugo combined automation with human expertise to improve the client's foundation models:

• Caption Review: Hugo identified errors, flagged inconsistencies, and rewrote captions to improve training datasets.

• Prompt Evaluation: The team assessed LLM-generated prompts and responses, refining them for clarity, accuracy, and intent.

• Performance Metrics: Hugo worked with the client to define key metrics such as response quality and evaluation criteria to align with business goals.

• Detailed Feedback: Each conversation turn was evaluated for tone, context, and reasoning, with clear rationales provided for preferred responses.

Results

Hugo's approach delivered measurable improvements:

• Achieved a 95% acceptance rate for evaluated captions and prompts.

• Reduced caption hallucinations by retraining the model to correct instances where objects were incorrectly identified in video frames.

• Improved grammar, consistency, and tone by reworking poorly structured sentences.

• Enhanced contextual understanding by incorporating North American expressions and references.

These improvements accelerated the client's validation process, improved customer experience, and led to additional projects with Hugo.