New blog post! 🚨 Avitan Gefen details a practical use case for LLMs in observability. So often we hear about what AI will be capable of in the future, but not too often do we hear the data science perspective from current projects. Our team has been working on using LLMs to “chat with data”–or form and execute custom database queries. “When it comes to leveraging LLMs for custom queries in observability, there are two main layers to consider. Layer One is the interaction between the user and the LLM. Layer Two is the interaction between the LLM and the data. Both layers have high degrees of complexity.” Check out the blog (link in the comments) to learn more about our process, learnings, and where we’re planning to take this next! #LLM #Observability
About us
Ready to meet the new brain of your observability stack? Get to know Senser, the AIOps platform that replaces the pain of monitoring and troubleshooting with insights you can actually do something with. Senser harnesses eBPF for immediate, zero-instrumentation visibility into production environments – and massively accelerates time to detect with ML-based insights into the root cause and business impact of service issues.
- Website
-
https://senser.tech
External link for Senser
- Industry
- Software Development
- Company size
- 11-50 employees
- Headquarters
- Ramat Gan
- Type
- Privately Held
- Founded
- 2021
Locations
-
Primary
Tuval St 11
Ramat Gan, IL
Employees at Senser
Updates
-
Senser’s own Avitan Gefen goes deep on how we’re using LLMs in a very targeted way to allow users to “chat” with their database. “When it comes to leveraging LLMs for custom queries in #observability, there are two main layers to consider. Layer One is the interaction between the user and the #LLM. Layer Two is the interaction between the LLM and the data.” With the right guardrails, we can create custom queries that deliver precise, actionable data without requiring users to know the database structure intimately. Curious to learn more? Check out the full blog on The New Stack for an in-depth rundown. 👇 https://lnkd.in/g26Zek9z
Excited to share that I have published another article in The New Stack on the role of LLMs for AIOps! Thank you to everyone who checked out the last one. Enjoy and let me know what you think! Check it out at the link: https://lnkd.in/dN7UkvQz
-
Super proud of this piece by our brilliant data science lead Avitan Gefen in The New Stack – documenting our approach to exploring use cases for #LLMs in #observability. “As innovators in the observability space, we are constantly updating our views and beliefs about emerging technologies. It is our passion and curiosity that lead us to experimenting with the latest and greatest, and we love sharing what we have learned with like-minded technologists.”
Thank you to The New Stack for featuring my recent article on how we're complementing the Senser AIOps platform with LLMs! https://lnkd.in/dCESr-au
Supercharge AIOps Efficiency With LLMs
https://thenewstack.io
-
Nice insights from Steven Vaughan-Nichols in The New Stack – “As organizations increasingly migrate their Kubernetes clusters to the cloud, lured by the promise of scalability and managed services, they all too often unwittingly open doors to sophisticated attackers.” Kubernetes clusters, while powerful and convenient, can introduce complexities that make it harder to detect anomalies. This creates a hotbed for potential security and performance issues. In cloud-native production environments, it’s always better to be proactive than reactive. Real-time #monitoring and #observability can keep potential attackers at bay, allowing you to focus on optimizing scale and performance instead of aimlessly hunting hidden threats. Senser provides cutting-edge tools that give you unparalleled visibility into your #Kubernetes environments, ensuring you stay ahead of potential threats. 🔗Link to the full article in the comments.
-
Good technical breakdown of the meltdown resulting from the latest Crowdstrike update (link in comments). Tl;dr #kerneldrivers open up all kinds of risks – including system degradation or even outages. When we built Senser, we were aware of these risks – from our own experience putting out fires as a result of errors in driver code. That's why we used #eBPF tech for lightweight, non-intrusive data collection. eBPF programs are executed in an isolated environment (so they can’t access or modify sensitive kernel data structures) and go through a verification process before they are loaded into the kernel. Of course comprehensive data collection is just the first step in smart observability. But today's meltdown shows the devastating cost of outages and the risks of kernel drivers – a good reminder of the benefits of safe, secure, and lightweight system monitoring.
-
-
Senser reposted this
Data science lead Avitan Gefen opens up on how we’re layering LLMs on top of Senser’s #aiops platform for automatic issue summarization and conversational root cause analysis. He shares great insights on: 💡Why LLMs are most powerful when coupled with #graphmachinelearning 💡Challenges we faced in generating a useful, reliable, and intuitive user experience 💡The most important use cases for LLMs in observability Link to post in comments below.
-
-
Data science lead Avitan Gefen opens up on how we’re layering LLMs on top of Senser’s #aiops platform for automatic issue summarization and conversational root cause analysis. He shares great insights on: 💡Why LLMs are most powerful when coupled with #graphmachinelearning 💡Challenges we faced in generating a useful, reliable, and intuitive user experience 💡The most important use cases for LLMs in observability Link to post in comments below.
-
-
Congrats Amir Krayden on the great piece in DevOps.com – highlighting why #LLMs alone are not enough for #rootcauseanalysis. “The process of converting structured, time-series data into meaningful text inputs for an LLM is far from trivial. It is the biggest bottleneck in using LLMs — transforming a map of your environment into specific and relevant training data that GenAI can use to generate insights beyond the generic (i.e., what would you get if you Google “why might a microservice fail?”).”
Thanks DevOps.com for publishing my latest piece on why #LLMs fall short for #rootcauseanlaysis! Some useful guidelines for leaders looking to harness AI for observability – plus why it’s hard to convert structured graph data into the type of context-rich input needed for LLMs. (All of it based on lots of sweat, tears, and lessons learned the hard way. 😅) https://lnkd.in/dzTYNSDC
Why Generic AI Models Fall Short for Root Cause Analysis - DevOps.com
https://devops.com
-
“The process of converting structured, time-series data into meaningful text inputs for an LLM is far from trivial. In fact, it’s the biggest bottleneck in using LLMs [for root cause analysis]: transforming a map of your environment into specific and relevant training data that genAI can use to generate insights beyond the generic.” In his most recent blog, Co-founder and CEO Amir Krayden dives deep into his perspective on why LLMs shouldn’t be used for root cause analysis. And, he reveals the “right way” to incorporate GenAI in observability workflows. #LLM #AI #GenAI #Observability
-
-
"What can soccer teach us about the impact of Graph ML on generating insights from complex, distributed systems? It turns out: a lot." Yuval Lev explains graph machine learning (#graphML) through a deep dive into Manchester United. Read on to learn what every SRE team needs to know about modeling interactions and dependencies in complex systems to reduce mean time to remediate (MTTR) service incidents. https://lnkd.in/gJJePhWi #SRE #Observability
Any day where I get to talk ⚽ is a good day… My co-founder Amir Krayden has talked before about the limitations of #largelanguagemodels (LLMs) for root cause analysis. Long story short: lack of context, inability to reason about dependencies, poor suitability for time-series data, and limited interpretability. So what’s a better fit? Graph machine learning (#GraphML). In my latest blog post, I demystify Graph ML by showing how it can track interactions and predict outcomes – on the soccer field. (Translation for all you Man United fans out there – football pitch.) A soccer team is a dynamic system with complex dependencies and multiple relevant environmental factors: a perfect fit for Graph ML. H/t 🎩 to the many researchers whose work laid the theoretical groundwork 📚 for this post. Give it a read to learn what soccer and Graph ML can teach us about root cause analysis in distributed production environments (see comments). 🤯