MinusX’s Post

I tried a demo of minusx.ai at the YC S24 kick-off retreat 3 weeks ago. And if I had this at Airbnb or any other job, I would never go back. to using a normal analytics UI. This is faster and easier, and for someone like me, where I either had to learn advanced SQL/R or rely on a data scientist for some research, this would have changed the game. Check it out: https://minusx.ai/

minusx.ai (YC S24) is a Chrome extension that adds a side chat to Jupyter and Metabase. Given an instruction, their agent operates your apps to analyze data and answer queries. Their AI data scientist is a scientist, not yet-another-new-analytics-platform. MinusX just works, in tools you already use and love. You can: 1. Generate hypotheses and explore data 2. Interop with MinusX to modify existing Jupyter notebooks or Metabase questions 3. Select a region and ask questions, or ask for modifications The lack of data analyst/data scientist bandwidth is a real pain. If you’re a programmer, you just want answers. If you’re a product manager, you just want answers. If you’re an analyst/scientist, you want 10 clones of yourself. Any new fancy “talk-to-data” platform requires you to migrate your data and convince your whole team to move over— just to start using it. MinusX solves all these and more, and generally gets out of the way. The founders, Vivek Aithal, Sreejith Puthanpurayil, and Arpit Saxena, have extensive experience in machine learning, data infrastructure, and systems engineering from their previous roles at companies like comma.ai, Meta, and Udaan. They strongly believe that you already have all the tools you need; you just need something like MinusX to use them. Congrats to the team on the launch! 🚀 https://lnkd.in/gNVYJijt

Did You Know? You don't need to be a data scientist to use Log Analyzer GPT. Its user friendly interface makes data analysis accessible to everyone. Try Now! https://lnkd.in/dH8nVaUe #LogAnalyzerGPT #AIInnovation #DataInsights #LogAnalysis #DevOps #AITool #LogFiles #DataAnalysis #DataAnalytics #TechTrends #AppDevelopers #SecuritySystem #AIRevolution #DataEngineers #MarketingProfessionals #GPT #GenerativeAI #ITProfessionals #IT #Marketers #DataDriven #LAGPT #DataDrivenSuccess

The Power of Data Analytics! At AQe Digital (Formerly AQe Group), we transformed a client’s operations by building an Integrated Analytics Platform for real-time insights, tailored to their unique needs. Using advanced tools like #Python, #Tableau, #AWS, and #PowerBI, we provided predictive insights and seamless user training, enabling them to address critical challenges like limited visibility into customer behavior and inefficient resource allocation. The results speak for themselves: 50% increase in customer retention, 35% improvement in operational efficiency, 25% faster time-to-market. Ready to transform your future? Partner with AQe Digital today! #DataAnalytics #BusinessGrowth #TechTransformation #data #aqedigital #datamanagement #gitex #business #success #datascience #ai #ml #artificialinteligence #machinelearning

I am thrilled to announce that my team Sedra Merkhan _ Ali Khaled Alawneh _ Ahmad Sadik and I recently completed a Kaggle competition focused on predicting diamond prices using regression models. This project was both challenging and rewarding, providing us with invaluable insights into data science and machine learning. 🚀💎 We began with thorough data cleaning and preprocessing, addressing missing values and outliers. We also engineered a new feature—diamond volume—to enhance our model's predictive power. Through exploratory data analysis, we visualized feature distributions and their correlations with the target variable. 📊🔍 We trained and evaluated several regression models, including Linear Regression, Lasso, Ridge, Decision Tree, XGBoost, AdaBoost, and Random Forest. The Random Forest Regressor emerged as the best-performing model, achieving the highest R-squared score. Further hyperparameter tuning using grid search and cross-validation significantly improved its performance. 🏆🤖 This experience underscored the importance of data quality, feature engineering, and model tuning. we are immensely grateful to our sponsoring company, SHAI For AI | شاي للذكاء الاصطناعي , for their support and the opportunity to advance our skills in this competition. 🌟🙌 For more details, please check out our project on GitHub: https://lnkd.in/dbP8HzjG #DataScience #MachineLearning #Kaggle #Teamwork #SHAILabs #DiamondPricePrediction

Data Engineers, this one's for you. A whole new dataverse to explore has opened up. As Large Language Models (LLMs) continue to speed up the data creation process, the need for efficient and scalable data processing has never been more critical. Dataverse is a new unified open-source Extract-Transform-Load (ETL) pipeline designed specifically for LLMs and with a focus on being user-friendly. It streamlines the entire ETL process, from retrieving raw data and preparing it for processing, to performing various transformations such as deduplication and cleaning, and finally loading the processed data into the desired storage destination. In addition to that, Dataverse is built with scalability in mind, seamlessly integrating with Spark and AWS EMR to handle data workloads of any size. And as an open-source library, it has the potential to become a central hub for LLM data processing, fostering collaboration and knowledge exchange within the AI community. ↓ Liked this post? Consider subscribing to my substack to support my work by following the link under my name 💡

I’m interested specifically in how the bias mitigation and toxicity filters work. I think we can all agree we don’t want bias or “toxicity”, but what determines toxicity or skew? Isn’t there a baseline perspective or “opinion” that serves as a guide to these aspects? I’d love to see / read more in depth on how Dataverse and others are handling this.

Tried to start a data project for more MYSQL eda practice. I downloaded a dataset from Kaggle. Didn't read the summary. 30 min later in EDA, realized all the counts and group by results were too balanced. I immediately noticed every useful business category was even. It dawned on me that it was fake AI generated data. I take pride in myself noticing it was fake from exploring it rather than knowing ahead of time. It's one skill to explore and find patters, it's another to realize you're looking at useless garbage.

GenAI has some of us talking about low-code analytics once again, but it's now new. At some point, the idea was to use drag-and-drop UI to enable non-technical users to work on analytics applications without using SQL or Python or having to engineer data at the code level. This was only partially successful because a clean data warehouse equals useful analytics applications, and low code couldn't solve data quality issues. In 2024, it's the same thing all over again. Low code just replaces or augments the drag-and-drop UI with a GenAI interface. What about data quality? Compute is, or often initially appears to be, so much cheaper now and that means data is being extracted and duplicated a lot more than before. This means more data to sift through for the average analytics engineer. Maybe low code should be about making this process easier. My experience is that there's been no investment in making it easier to understand raw data. In making sense of proprietary source system data at scale. Ironically, GenAI might have the answer. Educated guesses are often invaluable when you don't have domain expertise. GenAI offers access to a generalist who's great at understanding context and relies on a variety of expertise.