?? Building high-performing GenAI apps takes more than just good prompts. You need the right infrastructure, monitoring, and evaluation at every stage. Chase Fortier, our Head of Solutions Engineering, alongside AWS engineers, Naufal Mir and Jia You, will share what it takes to build and monitor?these systems on March 6th. Register ?? ?? https://lnkd.in/d6sP_2nF
关于我们
Comet is an end-to-end model evaluation platform built with developers in mind. Track and compare your training runs, log and evaluate your LLM responses, version your models and training data, and monitor your models in production — all in one platform.
Backed by thousands of users and multiple Fortune 100 companies, Comet provides insights and data to build better, more accurate AI models while improving productivity, collaboration, and visibility across teams.
- 网站
-
https://www.comet.com
Comet的外部链接
- 所属行业
- 软件开发
- 规模
- 51-200 人
- 总部
- New York，NY
- 类型
- 私人持股
- 创立
- 2017
- 领域
- Machine Learning、Data Science、Developer Tools和Software
产品
![点击此处查看Comet]()
Comet
数据科学与机器学习平台
Comet provides an end-to-end model evaluation platform for AI developers, with best-in-class LLM evaluations, experiment tracking and production monitoring. - Debug and evaluate your LLM applications with Opik - Track and visualize your training runs with Experiment Management - Monitor ML model performance in production with Production Monitoring - Store and manage your models with Model Registry - Create and version datasets with Artifacts The best part? Comet is free for individuals and academics!
地点
-
主要
获取路线
100 6th Ave
US，NY，New York，10013
Comet员工
动态
-
Amazing seeing Opik in action! ?? Great breakdown, Akshay Pachaar, we're excited to see more people exploring structured LLM evaluation! ??
I put Claude 3.7 Sonnet & OpenAI o3 in a code battle. . . Anthropic just dropped Claude 3.7 Sonnet and it gives SOTA performance on codegen! Today, we build a Streamlit app to compare and evaluate them using RAG over code (GitHub) Here's what you'll learn: - LlamaIndex for orchestration - Comet's Opik for evaluation - Streamlit for the UI The video below shows a quick demo of how it works! I have created a separate notebook for a formal evaluation task using Opik where you'll find all the code to create and load the evaluation dataset, run an experiment, create a custom evaluation metric and get the evaluation results in a dashboard. I have shared all the code in comments. ____ Find me → Akshay Pachaar ??? For more insights and tutorials on AI and Machine Learning!
-
?? Single-model evaluations can be biased, inconsistent, and costly—especially with large models like GPT-4o. LLM Juries offer a better alternative, using multiple smaller models to improve robustness and reduce bias at a fraction of the cost. ?? Inspired by traditional #ML ensembling, an LLM Jury has multiple models independently score an output, then aggregates their scores through a voting function. ?? ???????? ???? ?????????? ?????? ???????? ??????????????? Check out our new tutorial using Opik and OpenRouter: https://lnkd.in/e8d-DFG8 #GenerativeAI #ArtificialIntelligence #MachineLearning
-
?? Join the fourth edition of #ConvergenceConference on May 13-14, 2025! This two-day virtual event dives into GenAI Engineering: One Line at a Time. Register here for free ?? https://lnkd.in/dfRWtwe2
此处无法显示此内容
在领英 APP 中访问此内容等
-
??? Building a scalable Generative AI platform is challenging, but it doesn’t have to be. Join us and Amazon SageMaker for a technical session on: ? The importance of LLM observability in production ? How Comet’s Opik can track and monitor your LLMs ? Effortlessly setting up Comet within SageMaker AI Partner Apps ?? Thursday, March 6th | 13:00 - 14:00 EST ?? Register: https://lnkd.in/dHBztcRm
Streamline GenAI system evaluation and observability with Amazon SageMaker and Comet
aws-experience.com
-
???? Join a global community of developers on May 13-14th for Convergence 2025, a virtual conference dedicated to GenAI engineering. We'll explore: ?? The challenges of building and deploying LLM-based applications ?? Advanced LLM evaluation techniques ?? Responsible use of GenAI ??? Register for free:?https://lnkd.in/dfRWtwe2
-
-
? Opik officially has 5,000 GitHub Stars!?? Five months ago, we launched Opik out of a growing need from the community to be able to confidently test and trust their LLM applications. Since then, the adoption and engagement we’ve seen has been beyond what we could have imagined. ?? Opik trending on GitHub as the #2 top repo ?? Tens of thousands of users ?? Contributions and callouts from users like Andreas Nigg, Jeremy Mumford, Carlos Kemeny, PhDx2, and Prakash Chaudhary ?? Incredible projects powered by Opik, like Chia Jeng Yang’s PatientSeek, an open-source Med-Legal Deepseek reasoning model We're grateful for the entire community's contributions. Whether you’ve contributed code, shared feedback, or spread the word, we’re excited to keep building with you???
-
?? Proud to be a community sponsor at the AI Tinkerers – NYC x OpenAI Hackathon this weekend! ?? If you're attending, be sure to say hello to Claire L., who will be representing Comet and diving into our open‐source LLM Eval framework, Opik. Can't wait to see what teams build ??
AI Tinkerers is cooking this weekend around the world! ?? AI Tinkerers - NYC x OpenAI ?? AI Tinkerers - Paris x Anthropic ?? AI Tinkerers - Singapore x AWS ?? ?? ?? ?? ?? ?? ?? ?? ?? ?? ?? ??
-
Comet is now on Bluesky!??? As a team that's committed to investing in the open-source community, we're excited to join the conversation on Bluesky. ?? Come say hi, give us a follow, and join us as we continue to build. Find us at?https://lnkd.in/dDriMEY7 ??
-
-
Comet转发了
?? Opik Weekly Changelog ?? ?????????????????? ???? ?????? ????????: Multiple external contributions have been released this week ! Opik is gaining momentum even faster than I expected ! One of the best parts of working on Open-Source projects are community contributions, not only do they improve the overall features of the product but they also often improve the quality of the product significantly. From day one we decided to prioritize reviewing user contributions quickly and we couldn't be happier we did ! We also released: ? Performance improvements for workspaces with over 100 million traces ? Added support for cost tracking when using Gemini models ? Added diffing of prompt versions ? Improved support for Ragas metrics in `evaluate_*` functions in the SDK ? Added support for Bedrock `invoke_agent` API And as always, thank you to all of Opik's external contributors including Jeremy Mumford, Rahul Kadam, Prakash Chaudhary, @demdecuong and @jeffy !
-