Leading Gig Marketplace Monitors ML and LLM Applications Across Regions with Fiddler’s Unified AI Observability Platform
A Trusted Gig Marketplace Turns to AI Observability
This leading gig marketplace connects consumers with local professionals, allowing individuals and businesses to hire service providers for a wide range of tasks, including home improvement, events, wellness, business services, and among other types of services. The company must manage the differing needs of professionals and consumer audiences — ensuring strong partnerships with professionals while meeting consumer expectations — and has turned to AI applications to help navigate this challenge.
These applications include deploying an LLM-based chatbot to provide support for both professionals and consumers, as well as applications powered by traditional ML models for search engine rankings and calculating predicted customer lifetime value (LTV).
In order to confidently scale these services, the company partnered with Fiddler to monitor their models across LLM and ML applications.
Delivering Results Across AI Applications
With Fiddler, the gig marketplace achieved comprehensive visibility into their LLM and ML applications, with benefits including:
- Efficient Optimization: Customized metrics and charts aligned to business objectives enabled developers to more efficiently identify opportunities for model optimization.
- Improved Customer Service: Accelerated the investigation process, enabling faster and more accurate responses to customer complaints.
- Boosted Confidence: Increased trust in the performance and reliability of their production ML models and LLM applications across the organization.
- Reduced Risk: Reduced LLM hallucinations and safety risks, preserving brand reputation and enhancing user experience.
Overcoming Challenges to Enhance AI Performance and Security
Once implemented, the company faced several challenges in maintaining its AI solutions. Monitoring efforts for LLM applications and ML models were largely manual, requiring significant time and effort to debug and track performance issues — hindering their operational efficiency.
Additionally, key ML models, including those for customer lifetime value and search ranking use cases, needed improved accuracy to align with business objectives. Manually identifying areas of improvement was time consuming and costly.
This company was also concerned with protecting sensitive data, such as personally identifiable information (PII), from leaking through its LLM-based chatbot application. Without a scalable, unified ML and LLM monitoring platform in place, they were left vulnerable to potentially costly or reputation damaging incidents.
Driving Insights and AI Safety with the Fiddler AI Observability Platform
Since implementing Fiddler, the unified AI Observability platform for LLMOps and MLOps, has addressed these challenges by centralizing monitoring, and proactively surfacing deep insights into the company’s AI applications.
The company was able to optimize LLMs and ML models to improve chatbot interactions for vendors and end users alike, improving conversion rates and reducing operational costs.
Fiddler also enhanced their LLM application’s data security by detecting and alerting the company’s team about PII leakage, enabling them to iterate on the model and prevent risky incidents.
To learn more about how Fiddler’s AI Observability platform can help you leverage AI to innovate with confidence, book a demo, or read additional case studies.
