Observability

Byscientist
Close-up view of binoculars on a surface with sunset reflection in lenses.

Observability is important for AI and AI tools. It is the ability to monitor them for token usage, response quality and model drift. Typically, an AI system is monitored through logs, traces and metrics but an AI system on AI agent may need other metrics. Troubleshooting a complex AI system that produces its output probabilistically is much more difficult since subtle changes in the input layer may change the output and hence a very different observability metric is needed. The requirement may be simple – decrease the hallucinations, and latency while removing any observable bottlenecks.

Check out this AI observability video : https://www.youtube.com/watch?v=v_SPcIGOJvs

Similar Posts

  • Sparql tool – connect LLM to knowledge graphs

    Byscientist

    Talk to knowledge graph – sparql-tool Use LLM.: Hallucinations (sometimes good for creativity), Outdated knowledge, no access to your data ( trained on public knowledge). Solutions: Fine tuning – retrain the model on domain data RAG – Search: look things up in real time. Claude code looks at searched doc. Vector embeddings: semantic similarity search…

  • Text from Image

    Byscientist

    Starting with a base image, it is possible to generate a description. An AI (Qwen-3-VL) simple description : This is a simple, hand-drawn black-and-white sketch of a person’s upper body. Here’s a breakdown of the key elements: Hair: Long, flowing hair depicted with two smooth, curved lines extending from the top of the head down…

  • Explainable AI

    Byscientist

    A very traditional problem solving method is the following: given a set of features or variables, can we understand the features to form a conclusion. This could be something like a treatment strategy wherein the strategy is built on a series of data and then ingesting the data helps make a conclusion. However, an equally…

  • |

    Fitness landscape for antibodies (Flab2)

    Byscientist

    There have been many AI models to predict the structure of proteins, especially antibodies. There are several key aspects to developing antibodies as drugs beyond binding to the target of interest. These are : thermostability, expression, aggregation, binding affinity, pharmacokinetics, polyreactivity and immunogenicity. There have been many models available such as IgLM, ProGen2, Chai-1, ESM2,…

  • Error codes

    Byscientist

    There is no error code in the answers that are provided by AI prompts. It will return an answer that is the best fit to the prompt or the question, but it does not tell you the probability that it is not correct or that it is low probability of answer. The conversational AI will…