Vaibhav Jain

𝗜 𝗴𝗼𝘁 𝗳𝗼𝗼𝗹𝗲𝗱 𝗯𝘆 𝗺𝗶𝘀𝗹𝗲𝗮𝗱𝗶𝗻𝗴 𝗱𝗮𝘁𝗮 𝘁𝗿𝗲𝗻𝗱𝘀! Have you ever noticed patterns in data that seem to suggest one thing is causing another, when that's not really the case? Let me explain with an example. In the following chart I tracked two metrics for India from 1990-2020: forest area and mortality rates for kids under 5. Here's what the data shows: 1. Forest area in India has 𝗶𝗻𝗰𝗿𝗲𝗮𝘀𝗲𝗱 over the years🌳 2. Child mortality rates have 𝗱𝗲𝗰𝗿𝗲𝗮𝘀𝗲𝗱 🏥 At first glance, you might think - more trees must be leading to fewer kids dying! But that doesn't make much sense, does it? The truth is, these two trends are just correlated - they happen to move in opposite directions over time. But having more forests doesn't directly cause lower child deaths. There are likely other major factors involved, like: • Improved healthcare access • Better nutrition • Economic development So while the data shows a strong inverse correlation between forest area and child mortality, it's inaccurate to conclude that one is causing the other. 𝗖𝗼𝗿𝗿𝗲𝗹𝗮𝘁𝗶𝗼𝗻 𝗱𝗼𝗲𝘀𝗻'𝘁 𝗲𝗾𝘂𝗮𝗹 𝗰𝗮𝘂𝘀𝗮𝘁𝗶𝗼𝗻. 𝗧𝗵𝗲 𝗸𝗲𝘆 𝘁𝗮𝗸𝗲𝗮𝘄𝗮𝘆? 𝗔𝗹𝘄𝗮𝘆𝘀 𝘁𝗵𝗶𝗻𝗸 𝗰𝗿𝗶𝘁𝗶𝗰𝗮𝗹𝗹𝘆 𝗮𝗯𝗼𝘂𝘁 𝗱𝗮𝘁𝗮. Just because two trends co-exist, doesn't necessarily mean one is influencing the other. Dig deeper to understand the real underlying drivers. #dataanalytics #criticalthinking

10

"Continuous Learning" is like a 'Systematic Investment Plan', you will compound your benefits by remain invested, it will beat the market odds in the long run! It is most powerful tool which has potentially to get ahead of 'best performers'.

19

1 Comment

This a foundation course if you want to land up into the world of forecasting. I have done few forecasting projects earlier, but this course has exposed me to some lesser known fact essential if you want be ace time series modeling.

7

Your quest for data quality is actually a cry for better information design. Check out my latest article https://lnkd.in/gFDppRd3 where I delve into this topic. Special thanks to Tejas Chopra and Dan Jones for their valuable feedback. #DataQuality #InformationDesign #DigitalTransformation #AIfriendlyDataDesign

27

3 Comments

In one of the classes at Indian Statistical Institute, Kolkata Dr. Arnab Chakraborty shared a thought that stuck with me: "For a layperson, calculating the mean is simply finding the average, and the story ends there. But for a statistician, it’s the mean + e, a margin beyond which we as statisticians find the exceptional." This concept has resonated with me beyond statistics. Philosophically, I see the "mean" as where most of us fall, a baseline of skills, knowledge, and experiences that many share. But what sets me apart are the "anomalies," the intentional deviations, the unique margin of error I bring to my work It’s in these differences, my willingness to try new things, push boundaries, or take risks that I am able to distinguish myself. Like in data, the real insights and breakthroughs lie in these exceptions It is with confidence of accumulated successes, I have started embracing my unique error of margins and I find courage to stand out beyond average

508

20 Comments

I recall a time during the last decade when traditional Data Warehouses and bread and butter Data modeling were overshadowed by new age tech jargons. This was a time when Big Data Appliances were not ready for traditional BI apps based on normalised'ish denormalised schema like Dimensional Models. Normalisation became that infamous 'Bad-Boy' of tech world hated by everyone. I believe that narrative was created to buy time while the products were getting matured. The narrative created a shortage of such Data Modelers since majority invested their time and effort in navigating their FOMO of hi-tech jargons. Today, with the advent of Big Query, Redshift, Synapse Analytics, Snowflake etc. we are going back in time, the good'ol days of Dimensions and Facts. My message to the upcoming Data Engineers is to focus their effort on learning such Data Modeling concepts since this skill will take them far and to the top and enable persistent respect throughout their career. To put icing on the cake, Snowflake due to its crazy scalability of computational power is a flag bearer of Data Vault(Agile Data Modeling Pattern), a dreaded name in Data Modeling world (remember 4NF, 5NF, 6NF?) . I am enjoying this feeling of exclusivity 😀 . I would love to have a conversation if anyone is interested. #dataarchitecture #datamodeling #artificialIntelligence #datawarehousing #dataengineering #snowflake #bigquery #redshift #azuresynapse

14

Most of the #hiring #interview for #powerbi developers, a hiring manager will always discuss about #AI #visual in #powerbi. Keep yourself reday for same - 1. Key Influencers Visual • Purpose: To help users understand the factors that influence a specific metric or variable. • How It Works: This visual uses machine learning algorithms to analyze data and show which variables have the most significant impact on the selected outcome. For example, it can reveal the top factors that drive sales increases or customer satisfaction. • Use Case: It is valuable for data analysts and business users who want to identify trends and insights without needing to create complex statistical models manually. 2. Q&A Visual • Purpose: To allow users to ask questions in natural language and get answers in the form of visuals. • How It Works: The Q&A feature leverages AI to understand user questions and automatically generate a visual that answers them. For instance, if a user types “What were the total sales last month?” Power BI responds with an appropriate chart or card displaying the result. • Use Case: This tool is helpful for quickly gaining insights without building a dashboard from scratch, making it user-friendly for non-technical team members. 3. Decomposition Tree Visual • Purpose: To break down and analyze complex data hierarchies. • How It Works: The decomposition tree uses AI to suggest high-impact areas and provide a detailed breakdown of data points. Users can click on different branches to explore data, with Power BI automatically indicating which categories contribute most significantly to the overall metric. • Use Case: Ideal for root cause analysis and understanding contributions from different data segments, such as analyzing why revenue has increased or decreased. These AI-powered visuals enhance the analytical capabilities of Power BI, making it easier for users to draw actionable insights from complex data.

44

4 Comments

Hi All, In continuation of the last webinar ,we are scheduling our next webinar on Sunday 29th Sep @9:00 PM(IST) We will be sharing the details about the Data Science /AI/ML course in details ,which we are going to launch soon along with timelines. Anyone looking for transition into Data Science/AI/ML ,i invite them to attend this webinar. Hurry up and register. *Its FREE* https://lnkd.in/gWBncT99 Follow this link to join my WhatsApp group for Data Science /AI/ML/GenAI updates: https://lnkd.in/gkn9iFXe This is High Level Flow of our Plan which we will discuss in detail in webinar. Python-->Statistics-->Machine Learning-->Time series-->Deep Learning-->NLP-->Gen_AI #Python #DataScience #ML #DL #NLP #GenAI #KnowledgeSharing

8

1 Comment

I'm very impressed with #llama3 . I tried a number counting task. It started with a very detailed Chain of Thought attempt at answering it. It got the answer wrong at first. I followed up and asked it to *reflect* on its answer and evaluate it's own answer (I didn't say that it was wrong and needed to change, just asked it to reflect)... and it corrected itself! Self-reflection and adaptation is something I see in models the size of GPT4, not in an 8b model. This is very impressive. Plus, I love that Llama3 has a personality. It's not a dry robot like many others. USER: The following is a multiple choice question about high school mathematics. Answer to the best of your ability. How many numbers in the list 25,26, ..., 100? (A) 75 (B) 76 (C) 22 (D) 23 AI: Let's dive into it! To find the answer, we can count the number of consecutive integers from 25 to 100: 25, 26, ..., 100 = 25 + x, where x is an integer The first term is 25, and the last term is 100. To find the common difference (d), subtract the first term from the second term: d = 26 - 25 = 1 Now, we can use the formula for the nth term of an arithmetic sequence: an = a1 + (n-1)d where an is the nth term, a1 is the first term, n is the term number, and d is the common difference. In this case, the last term (100) is equal to the sum of the first term (25), the common difference (1), and the term number (x): 100 = 25 + x - 1 Simplifying the equation: 75 = x So, there are 75 numbers in the list from 25 to 100. Comparing our answer to the options: (A) 75: Correct! (B) 76: Incorrect (C) 22: Incorrect (D) 23: Incorrect Therefore, the correct answer is (A) 75. USER Can you reflect on your answer and evaluate it? AI I'd be happy to reflect on my previous answer! Upon reviewing my response, I notice that I provided a detailed explanation for finding the number of consecutive integers in the given list. However, upon re-evaluating the options, I realize that I didn't actually provide evidence to support option (A) 75 being correct. In reality, the correct calculation would be: 100 - 25 + 1 = 76 There are indeed 76 numbers in the list from 25 to 100. I apologize for any confusion caused by my earlier response. In hindsight, a more accurate answer would have been to acknowledge that the correct answer is not among the options provided and offer an explanation of how to calculate the correct number of consecutive integers. Thank you for giving me the opportunity to reflect on my previous answer!

24

🤔 Not sure which career to choose in India? Let's break it down with some of your favorite movie heroes! 🌟 🎯 Data Science: Be the spy like in Gudachari, using data to solve mysteries and predict trends! 📊 🔧 Mechanical Engineering: Just like Pushpa, build and fix things to keep our world moving! 🚗 🤖 Artificial Intelligence: Embrace futuristic tech like in Kalki 2898 AD, creating smart systems! 🌐 🛡️ Cybersecurity: Protect the digital world like Kamal Hassan in Vikram! 🛡️ Choose a path that excites you and matches your skills. The future is yours to shape! 💪 Follow us for more career tips! 🚀 Watch Now #CareerChoices #DataScience #MechanicalEngineering #ArtificialIntelligence #Cybersecurity #Gudachari #Pushpa #Kalki2898AD #Vikram #CareerTips #India

2

Hi #datafam, I got a chance to discuss the multi-dimensional multi-level set action in Tableau Buddy: TUG Thank you Avinash Reddy Munnangi and team for organising this session #TUG #Tableau #tableaubuddy

16

1 Comment

"𝐉𝐨 𝐝𝐢𝐤𝐡𝐭𝐚 𝐡𝐚𝐢 𝐰𝐚𝐡𝐢 𝐛𝐢𝐤𝐭𝐚 𝐡𝐚𝐢, 𝐛𝐨𝐬𝐬........" 🙂 Don't believe me? Let me prove it to you. ⤵ This exact strategy helped my mentee, Rashi Agarwal land her dream job at the top product-based company ! When I first spoke with her, she had completed 7 certifications and 6 projects to become a data analyst. Despite all her hard work and technical skills, she struggled to get job opportunities for over 2 1/2 months. The main reason? ⤵ She couldn't showcase her skills effectively to recruiters and hiring managers. Plus, she had a 3-year career gap. Crazy, right? 😲 But we took on the challenge, and in the next 8.5 𝐰𝐞𝐞𝐤𝐬, Rashi Agarwal secured a 𝐃𝐚𝐭𝐚 𝐀𝐧𝐚𝐥𝐲𝐬𝐭 position at S&P Global, one of the top 10 firms in the data domain space. 𝐇𝐞𝐫𝐞'𝐬 𝐭𝐡𝐞 𝐞𝐱𝐚𝐜𝐭 𝐩𝐫𝐨𝐜𝐞𝐬𝐬 𝐰𝐞 𝐟𝐨𝐥𝐥𝐨𝐰𝐞𝐝 𝐭𝐨 𝐚𝐜𝐡𝐢𝐞𝐯𝐞 𝐭𝐡𝐢𝐬: 🎯 Optimized her LinkedIn to attract attention. 🎯 Coached her on a robust networking strategy. 🎯 Held mock interviews to prep for every scenario. 🎯 Worked on her LinkedIn branding to position herself as an expert. 🎯 Discussed hidden strategies to negotiate a salary better than market standards. Rashi Agarwal consistency and dedication paid off, and I'm incredibly proud of her achievement. 🙌 I hope her story inspires others to restart their careers, even if they've faced setbacks. 𝐖𝐚𝐭𝐜𝐡 𝐡𝐞𝐫 𝐂𝐨𝐦𝐩𝐥𝐞𝐭𝐞 𝐉𝐨𝐮𝐫𝐧𝐞𝐲 𝐭𝐨 𝐠𝐞𝐭 𝐚𝐜𝐭𝐢𝐨𝐧𝐚𝐛𝐥𝐞 𝐈𝐧𝐬𝐢𝐠𝐡𝐭𝐬 𝐨𝐧 𝐭𝐡𝐢𝐬 𝐏𝐨𝐝𝐜𝐚𝐬𝐭 : https://lnkd.in/gYJ8xyFV To all the aspiring data analysts aiming to land their dream job, share and repost this. If you are an 𝐈𝐓 𝐩𝐫𝐨𝐟𝐞𝐬𝐬𝐢𝐨𝐧𝐚𝐥 with 1-5 years of experience or a graduate with technical skills struggling to land a job, book a 1-1 𝐜𝐨𝐧𝐬𝐮𝐥𝐭𝐚𝐭𝐢𝐨𝐧 𝐜𝐚𝐥𝐥 from the Featured or comment section to land your dream job.

11

4 Comments

Something I've written on how a T20 batting innings can be assessed. This framework can be extended to assessment of individual players as well, which I will do in the susequent posts. For now, please feel free to comment and if you like the content, please subscribe. #cricket #cricketanalytics #data #dashboards #datavisualization #sportsanalytics https://lnkd.in/gb7T-UzJ

28

6 Comments

If anything, the #gartner #magic quadrant for data science and machine learning platforms is a nostalgic account for bringing back good old times…… I am always surprised how each year Gartner can compare tools which are so vastly different in many aspects. But I am really impressed that the 2024 quadrant still shows some #ancient_tools of the past that are hardly used by anyone today for data science! But to me personally, these tools bring back very good memories. 👉 Mathworks Matlab I did my graduation project in 1995 with matlab at Shell…… 👉 IBM SPSS I was using this in Brussels for a project. This was in 2005, way before remote working. I had to be there in hotels….. 👉 SAS My first sas base data step in 1999 for ABNAMRO impressed my manager…… #gartner #ancient #relics from the past.

23

5 Comments

A must-watch interview to understand AI Reasoning, by Prof Subbarao Kambhampati. He explains how LLMs say things in the deductive closure of base facts and make us believe that they reason, while the real reason behind this reasoning capability is that the training data has both base facts and what you can derive based on them. He shares how a planning dataset called Blocks World gets near 0% accuracy with the latest LLMs when you make changes to the dataset that would not affect the performance of a logical system. The neural and symbolic AI systems are both going to be important in the future of AI. #ai #future

15

5 Comments

Hello everyone, please read my most recent blog, "The Role of Double Exponential (Laplace) Distribution in Optimizing AI Models". Do not forget to like, share, and subscribe to stay updated. #statistics #statisticsblogs #datascienceblogs #probability

1

RAG 101 Retrieval-Augmented Generation (RAG) is the process of optimizing the output of a large language model, so it references an authoritative knowledge base outside of its training data sources before generating a response. I am sure there is absolutely no one in my circle who has not heard about RAG. But are you following all the best practices? If you are starting to implement/learn RAG , then start with the titles in the box. 1. Why we need query classification 2. What are the retrieval methods 3. What is reranking 4. Why do we need to chunk data 5. What are embedding models used for 6. Why do we need vector database 7. What is document repacking 8. What happens in summarization 9. What is generator and retriever fine tuning 10. What are the evaluation strategies. If you can answer all the these 10 questions , then we go to the specific techniques for doing each of these 10 steps. Paper : https://lnkd.in/gYBVrc9d #generativeai #rag #llm

46

2 Comments