Activity
4K followers
-
Viraj Pawar shared thisViraj Pawar, congratulations on completing Knowledge Graphs for AI Agent: API Discovery!Viraj Pawar, congratulations on completing Knowledge Graphs for AI Agent: API Discovery!
-
Viraj Pawar shared thisViraj Pawar, congratulations on completing Attention in Transformers: Concepts and Code in PyTorch!Viraj Pawar, congratulations on completing Attention in Transformers: Concepts and Code in PyTorch!
-
-
-
Viraj Pawar liked thisViraj Pawar liked thisStatQuest makes me unreasonably happy… and now I know why 💥🦖(it's the nerd joy 🤓) You know that *rare* feeling when learning doesn’t feel like effort… but like opening a comic book you don’t want to put down? Yeah. That. Today, I added the **Statistics essence (right in the center)** to my collection — and suddenly everything feels so satisfying. But wait… this isn’t just about books. This is about the crew 🔴"🟠"🔵 **StatSquatch** — making probabilities feel like a snack decision 🍟 **Normalsaurus (Norm)** — casually running experiments like it’s a hobby 🧪💥🦖 **The Gamma Monster** — that wise senior who *actually* explains things clearly 🧠 Tell me honestly… Who knew statistics would come with *characters* AND personality?! Somewhere between dinosaurs, monsters, and fries… I stopped being intimidated by data. And started enjoying it. Like… genuinely smiling while learning kind of enjoying 😄 And the newest addition.... It’s the **glue**. The “ohhh NOW this makes sense” moment. **StatQuest didn’t just teach me concepts.** It tricked me into liking them. And that’s dangerous… because now I want to learn *more*. TRIPLE BAM!!! Joshua Starmer PhD ( I intentially put up the screen from one of your lesson, just to show the degree of appreciation) #StatQuest #TripleBam #Statistics #MachineLearning #AI #LearningCanBeFun #NerdJoy #DataScience #GrowthMindset
-
Viraj Pawar liked thisViraj Pawar liked thisTime to test what state of art AI Agents are capable of! Over the next few days, I'll be building a proof of concept for an AI Agent tackling one of the most high-stakes problems in financial compliance: OFAC sanctions screening and will be documenting my journey here. Dataset - 18,714 individuals and sanctioned entities from the official OFAC list. Problem - Currently OFAC severely punishes any business (in millions!) who transacts with people/entities in sanctioned lists knowingly or unknowingly. This leads to several millions of loss per year draining their profit margins thin. Bottleneck - There is currently NO github or big business who has Implemented a AI Agent for this use case. I will be handling this challenge head on purely with deep reasoning capabilities of claude, a lot of coffee + research all day. Approach - Theorycraft a fuzzy algorithm that will cast a wide layer and then use AI Agents to optimize the output. My first goal is to accurately determine which combinations of algorithms to use to implement fuzzy matching. #MachineLearning #AI #AIagents #Claude #ML #AIEngineer
-
Viraj Pawar reacted on thisViraj Pawar reacted on thisHow much artificial brain does it take to control a converter? We ran an interesting experiment: coupling a Reinforcement Learning algorithm from a Python AI library with a circuit simulator. We provided only a netlist of the converter topology – no controller, no prior information. The AI had no idea what it was dealing with (no training data!) and had to figure things out by trial and error, just like a student new to power electronics: opening and closing switches, watching currents and voltages. We tasked the AI with controlling sinusoidal current (as required by legal standards) while keeping losses small (otherwise the converter might literally go up in flames). And it worked amazingly well! The AI discovered a complete controller structure with optimized parameters. Now we are gradually increasing the converter’s complexity. We will present this work at IECON 2025 in Madrid, Spain, and ICRERA 2025 in Vienna, Austria. Below you can see the converter with its “brain” in the form of a Deep Neural Network (DNN) – and a tardigrade, a tiny animal with a brain of roughly comparable size.
-
Viraj Pawar liked thisViraj Pawar liked thisWell, that was fun! 🤩 On Tuesday this week, Ascent hosted the second GenAI Meetup in Graz, organized by Ruben Hetfleisch and Marc Hollomey, MSc, MBA from AI Austria. Again, it was a great combination of scientific and project-related presentations where I had the chance to be one of the speakers and share the stage with the awesome René Berndt, Malte Scheuvens and Mathias Ptacek 🚀. Looking forward to the next event to enjoy insightful talks and meet the growing AI community Graz! Also a huge thank you to our team at Ascent for turning our office into an event location! Pic by Jelena Z. 📸
-
Viraj Pawar liked thisViraj Pawar liked thisWe just wrapped another insightful session at the RHI Magnesita Data Science Club focused on AI and Machine Learning in R&D. Here are the core highlights: 🔬 AI in Laser Tech Prof. Dr. Carlo Holly and Dr. Cord Fricke-Begemann shared how AI enhances laser manufacturing, from real-time monitoring to smart beam shaping using optical neural networks. Thank you again for joining it! ♻️ Smarter Recycling In the EU-funded RESOURCE project, AI is leveraged to autonomously sort refractory materials, reducing waste and enabling circular value. 🤖 GenAI as a Daily Teammate Alexander Leitner showed how GenAI tools (like ChatGPT) support R&D, coding, presenting, troubleshooting, as well as boosting productivity and creativity. Thank you as well for your presentation, Alexander! 💡 Key Takeaways ✅ AI isn’t just a tool, it’s a strategic co-pilot in R&D ✅ Data quality and domain knowledge are critical enablers ✅ Even non-experts can leverage GenAI to automate, explain, and explore ✅ AI accelerates experimentation, improves accuracy, and unlocks circular economy value #RHIM #AI #DataScience #Innovation #RND #Sustainability #Recycling #GenAI #LaserTech #MachineLearning #DigitalTransformation
-
Viraj Pawar liked thisViraj Pawar liked thisHey, dashboard designers. There are only *five* ways to use color in data visualization, but this is where most people mess up... big time. Hello! I’m Steve Wexler, founder, principal, and chief chart looker-atter at Data Revelations. I am one of three authors of The Big Book of Dashboards and the author of The Big Picture. More importantly I’m here to make sure you create dashboards that are truly useful. AND... I want to make sure you don't mess up with color! Join me for my upcoming dashboard design workshop online May 27 and 29 (see link in comments). Note: Early-bird pricing ends April 27.
-
Viraj Pawar reacted on thisViraj Pawar reacted on thisA gentle reminder for Data Scientists (you'll thank me later) Learn Statistics before DL Learn Pandas before PySpark Learn Modeling before MLOps Learn ARIMA before TimeGPT Learn Numpy before TensorFlow Learn Simple Models before Ensembles Learn Linear Algebra before Neural Nets Learn One-Hot Encoding before Embeddings Learn Linear Regression before Transformers Learn Data Cleaning before Feature Engineering Remember: Basics First 🔥 𝗠𝗼𝗿𝗲 𝗣𝗿𝗮𝗰𝘁𝗶𝗰𝗮𝗹 𝗠𝗟 𝗖𝗮𝗿𝗲𝗲𝗿 𝗔𝗱𝘃𝗶𝗰𝗲 from me here: https://lnkd.in/emb4cFCS P.S: Inspired by Neo Kim
Experience & Education
-
Freelance
View Viraj’s full experience
See their title, tenure and more.
By clicking Continue to join or sign in, you agree to LinkedIn’s User Agreement, Privacy Policy, and Cookie Policy.
Licenses & Certifications
-
Structuring Machine Learning Projects
DeepLearning.AI
-
-
-
-
-
-
-
-
German Language A1.1
Goethe-Institut Irland
Issued -
Volunteer Experience
-
Computer tutor
Age Action Ireland
- 5 months
Education
Teach basic computer skills to people over the age of 55 under Age Action's Getting Started Computer Training programme.
Visit libraries, community centres, schools, colleges, family resource centres, corporate offices, and housing complexes for training for older people.
Courses
-
Advanced Embedded Systems
-
-
C# programming
-
-
Microsoft ASP.NET
-
Projects
-
Car damage and severity detection using CNN
-
-
Home and work location detection using Density-based spatial clustering of applications with noise clustering.
-
-
Driver identification using a statistical model
-
-
Driver risk assessment using a statistical model
-
-
Speech utterance classification for triage of 911 emergency calls
-
See projectMotivation: No automated system to prioritize 911 emergency based on the nature of emergency of the caller.
Problems: High call prioritization time for a real emergency.
Solution: Determine priority of calls before received using the speech utterance of the caller.
Techniques: Natural Language Processing, Text mining, Supervised learning methods.
Data Source: 911 audio calls.
Pre-processing: Speech to text API utilization (IBM, Google, CMU), text preprocessing.
Feature…Motivation: No automated system to prioritize 911 emergency based on the nature of emergency of the caller.
Problems: High call prioritization time for a real emergency.
Solution: Determine priority of calls before received using the speech utterance of the caller.
Techniques: Natural Language Processing, Text mining, Supervised learning methods.
Data Source: 911 audio calls.
Pre-processing: Speech to text API utilization (IBM, Google, CMU), text preprocessing.
Feature extraction: Bag-of-words, TF-IDF, Word2vec
Classification models: Support Vector Machine, Logistic Regression, Naive Bayes, AdaBoost.
Results: Accuracy achieved for 1. SVM - 78%, 2. AdaBoost - 75%, 3.Naive Bayes- 73%, Logistic Regression- 59%.
Conclusion: A classification model can be used with SVM classifier and bag-of-words features for triage of 911 calls.
Future work: More data required to get better accuracy. Utilize Neural networks for classification.
Honors & Awards
-
Microsoft AI Hackathon
Microsoft
• Participated in Microsoft Azure AI hackathon with Telematicus Data Science team to create a prototype product using Azure AI services.
• Create a statistical model for vehicle maintenance prediction using Azure machine learning server dependencies.
• Integrated Azure LUIS (Language Understanding) service with existing Telematicus shack framework to create a chatbot prototype for vehicle maintenance prediction.
-
Microsoft AI Hackathon
Microsoft
• Participated in Microsoft Azure AI hackathon with Telematicus Data Science team to create a prototype product using Azure AI services.
• Utilized Azure Machine Learning studio for performing exploratory data analysis on GPS data for driver behavior classification.
• Explored different Microsoft Azure AI services for prototyping statistical model for driver behavior classification.
Languages
-
English
Full professional proficiency
-
Hindi
Native or bilingual proficiency
-
German
Elementary proficiency
Other similar profiles
Explore more posts
Explore collaborative articles
We’re unlocking community knowledge in a new way. Experts add insights directly into each article, started with the help of AI.
Explore More