How to trust LLM against hallucinations using langkit & whylogs

Soniya Rangnani (~soniya3)




  • Large Language Model (LLM) adoption is reaching another level in 2024. While the size and capabilities of large language models have drastically increased over the past chouple of years, so too has the concerns imprinted into these models. Before AI based product is deployed, the correct methods and metrics must be implemented to ensure that the different dimensions of bias are captured in LLM outputs. In particular - this can be caused from retrieving irrelevant context, not enough context, and more.
    • Python Libraries- Langkit and whylogs can solve this problem. It can sit as the evaluation layer for the LLM stack, allowing you to shorten the feedback loop and iterate on your LLM app faster. This talk will cover all of the methods which are used to either detect, estimate, or filter out hallucination outputs in LLMs. Examples of metrics include accuracy, sentiment, fairness, and more.
    • This talk is for data scientists, engineers, leaders and whoever is interested in AI related applications Key Takeaways:
    • Understand common failure for LLM apps
    • Understand different evaluations that are useful for reducing hallucination, improving retrieval quality & more.
    • Understand how to introduce langkit and whylogs for safeguarding LLM’s output Talk outline:
    • Prompt engineering & challenges 2 minutes
    • Hallucination reasons & pitfalls 8 minutes
    • Introducing WhyLogs 5 minutes
    • Mitigation Techniques 5 minutes


Basics of Web App & LLM/ChatGPT.

Video URL:

Speaker Info:

  • Soniya Rangnani is Senior Data Scientist of 8 years of experience in AI/ML field. She has worked for reputed companies like Microsoft .
  • She is published author for International ML Conferences & alumni of Indian Institute of Science, Bangalore.
  • he is currently working with US Startup PriceLabs; working on relevant AI/ML problems like dynamic price recommendation.
  • She has successfully designed, built & maintained many ML systems in the space of information retrieval systems like recommendation systems, ad-ranking model, NLP text matching algorithms, title-text generation using LLMs on cloud-based tech stack.
  • She is active member of GDG Pune & WTM communities

    Profile link:

Speaker Links:

Section: Artificial Intelligence and Machine Learning
Type: Talk
Target Audience: Intermediate
Last Updated: