Tushar Mudgal

Attended “Build a RAG App with Redis” — a hands-on AI workshop. Covered practical topics like data ingestion, vector search choices for large-scale datasets, semantic caching for performance and cost optimization, routing patterns, and PII best practices. The session helped clarify trade-offs in search strategies and how caching + routing can significantly improve latency and reliability in RAG systems. Good insights into designing AI applications with production constraints in mind.

1

🚀 AI Powered Stock Market Forecasting System | End-to-end data science project Just uploaded the first walkthrough video of my new project - an AI powered stock prediction system built from scratch. I kept this project simple but powerful: 🔧 Custom API for market and news data 🤖 ML models for price forecasting 📰 NLP model for news mood 🎨 Backlit user interface for pure visualization Why am I making this? Well... partly to challenge myself and partly because maybe, just maybe, an HR person will look at this and think: "Hmm…this guy is really solid. Let's talk to him." I'm proud of how far I've come in Python, ML, NLP and computer engineering but still learning every day. If you find this project interesting, I would love your feedback, support or even a like. Thanks for reading - more videos coming soon! #datascience #machinelearning #python #ai #nlp #stockmarket #streamlight #opentowork

17

1 Comment

Is partitioning data overrated when scaling ML inference on Databricks? A recent case study compared two approaches – Liquid tables versus traditional partitioned tables – and here is what stood out: Key Takeaways • Liquid tables delivered up to 4x higher concurrency with lower latency • Partitioned tables benefited from data skipping but suffered from small-file overhead • Adding a salt key mitigated data skew but introduced complexity and planning overhead • Autoscaling clusters ensured cost efficiency without sacrificing performance • Optimal file size (100-200 MB) mattered more than deep partition hierarchies Controversial Insight Avoid over-partitioning. Too many partitions can throttle the Spark driver and increase job latency. Sometimes simpler Liquid tables win. Actionable Recommendations 1 Start with Liquid tables for ease of management and predictable performance 2 Use autoscaling clusters to balance cost and throughput 3 Aim for medium-sized files instead of exhaustive partition levels 4 Introduce salting only when you hit severe data skew, not by default By challenging the assumption that more partitions always equals better performance, you can streamline your MLOps pipeline and reduce costs. #MachineLearning #MLOps #Databricks #DataEngineering #BigData #Analytics

1

Learn the crucial difference between platform and analytics engineers for a successful data platform. By Ashok Singamaneni and Databricks' Gaurav Nanda

2

Having worked with a wide variety of LLMs, this is a great roadmap to follow for anyone who wants to truly learn NLP

9

Enterprise AI infra startup Kluisz.ai raises $9.6 Mn in seed round led by RTP Global ◾The founding team includes former Oyo COO and CPO Abhinav Sinha, ex-McKinsey partner Vamshidhar Reddy, and former Jio cloud executive Abhijeet Singh. https://lnkd.in/gB9EJuwr

2

Sarvam-M: A 24B Parameter Multilingual AI Model is finetuned Mistral-Small with Indic language desi nuances covering 11 major languages from india 🇮🇳 Key Features: • Dual-mode interface: Quick "non-think" responses + detailed "think" mode for complex reasoning 🧠 • Strong performance on math (GSM-8K) and coding (SWE-Bench) benchmarks 📊 • Supports both native Indic scripts and romanized text ✍️ • Built for real-world multilingual conversational agents 💬 I think it's a good start , instead of training model from scratch Sarvam focused on building indic dataset and model like in video shows good translingual conversational capability . Try it out via the Colab link - https://lnkd.in/gzcW4hru Playground - https://lnkd.in/gXijZrkY

27

1 Comment

Tackling Massive Data with PySpark Internals Have you ever struggled with processing huge datasets efficiently? PySpark has been my go-to solution for handling large-scale data processing with ease. Here’s why PySpark is a game-changer: Distributed Computing: Leverages Spark’s in-memory processing to handle terabytes (or even petabytes) of data. Optimized Execution: The Catalyst Optimizer and Tungsten Engine ensure queries run at high speed. Scalability: Seamlessly scales from a single machine to a massive cluster. Rich APIs: DataFrames, SQL, and RDDs provide flexibility for different use cases. My Approach: When working with massive datasets, I focus on: Partitioning data wisely to avoid skew and improve parallelism. Caching frequently used datasets to minimize recomputation. Optimizing Joins & Aggregations to reduce shuffles and I/O overhead. Monitoring Spark UI to identify bottlenecks and fine-tune performance. Pro Tip: Always check your execution plan (df.explain())—it reveals hidden inefficiencies. Have you worked with PySpark on large-scale data? What strategies do you use? Let’s discuss in the comments. #PySpark #BigData #DataEngineering #ApacheSpark #Optimization #DataProcessing

1

FractalX Blog: https://lnkd.in/geSGJdHq — A deep dive into static AST analysis and build-time code generation - how FractalX automates the decomposition of Spring Boot monoliths into production-ready microservices. #fractalx #springboot #software_engineering #lka

9

Before LLMs became mainstream, the groundwork was already in motion. Abhay Tandon talks about running experiments back in 2017 while working at Target. The approach was simple but powerful: Use NLP to scan massive datasets. Identify the data points that actually matter. Extract signal from noise. Then use NLG to auto-generate reports. It wasn’t hype. It was structured automation. The core idea, though, hasn’t changed: 🔺 Automation scales analysis 🔺 Context increases decision quality 🔺 Generative AI compounds structured thinking AI didn’t suddenly appear in 2022. The foundation was always there. The tools just got exponentially better. 👉 Subscribe for more AI, data, and founder insights: https://lnkd.in/g3YQwF6h Ayush Bothra Shruti Sharma #AI #GenerativeAI #NLP #LLM #DataAnalytics #Automation #FounderJourney #InnovatorsAndDisruptors

8

🚀 Batch Telemetry Agent in Action: Estimate timings or completion timings Insights for a Batch Job! Ever wondered how to estimate batch job timings without complex ML models? Here’s how our Batch Telemetry Agent delivers business-ready predictions using a heuristic-based approach: ✅ Complete Documentation Generated Automatically ✔ Executive Summary ✔ Critical Issues & Recommendations ✔ Ready-to-use KQL Query ✔ Role-specific guidance for Dev, Ops, and Business 🎯 Estimation Logic Explained The query uses a Statistical Baseline with Multiplicative Adjustment Factors approach. It's NOT a machine learning model, but rather a heuristic-based prediction system that combines historical statistics with real-time system conditions. 📊 Prediction (Estimate) Logic in a Nutshell Key Steps: 1️⃣ Baseline: Use median (P50) for stability, not average. 2️⃣ Multipliers: Adjust for system health: Throttling: Up to 2× slower if >50% throttled CPU: Up to 1.5× slower if >80% usage Threads: <2 threads → 1.8× slower (your critical issue) Queue: Large backlog → up to 1.4× slower 3️⃣ Combine: Multiplicative model (factors compound). 4️⃣ Confidence: P75 & P95 bounds for conservative/pessimistic estimates. Classification: ✅ Rule-based, explainable, real-time adaptive ❌ Not ML, no training, no trend modeling NOTE: This feature will come up with Version 3.0 of Batch Telemetry Agent. #Dynamics365 #MicrosoftCopilot #AIinERP #Observability #Telemetry #AIAgent #CopilotStudio #MicrosoftFabric #DigitalTransformation #IntelligentAutomation #ProactiveMonitoring #ApplicationInsights #D365Finance #D365SupplyChain #ERPInnovation •📥 Link to download the agent: https://lnkd.in/gyMhnnV2 •See Batch Telemetry in Action: https://lnkd.in/gpnY6p4E •Join Viva Group: https://lnkd.in/gxi-Nw5x

53

1 Comment

Java Capital Launches Rs 400 Crore Fund for Seed-Stage Deeptech Startups Early-stage venture capital firm Java Capital has launched a Rs 400 crore fund focused on supporting seed-stage deeptech startups building strong, IP-led technology businesses. Read More: https://lnkd.in/gigF2u9F #productdevelopmentcycles #Indiasdeeptechecosystem #globalmarkets #deeptechstartups

5

Bhavin Turakhia didn’t chase hype. He built profits. 🚀 Started at 17, built Directi created global SaaS products like Radix , Flock & Zeta and crossed $1B+ in exits — mostly bootstrapped. Key Learnings: • Profitability beats valuation • Build global, think long-term • Discipline > funding • You don’t need noise to win 📌 Quiet execution builds loud results. #BhavinTurakhia #StartupLessons #Bootstrapped #SaaS #Entrepreneurship

6

🎉 Today’s AI Learning Update I delved into the renowned research paper “Attention Is All You Need,” a pivotal work that introduced Transformers 🔥. Here are the key insights I've gathered: 🔹 Pre-Transformers era saw challenges with slow sequence models like RNNs & CNNs due to their sequential or chunk-based text processing methods. 🔹 The breakthrough concept proposed discarding recurrence and convolution in favor of universal attention. 🔹 Attention facilitates direct connections between every word in a sentence, enabling seamless interactions like "The book … was amazing.” 🔹 The inception of the Transformer architecture featured: - An Encoder–Decoder framework. - Self-attention for word interactions. - Multi-head attention for diverse interpretations. - Feed-forward layers & residual connections for stability. 💡 Significance: Transformers offer: - Accelerated training through parallelization. - Enhanced handling of distant dependencies. - Pioneering translation outcomes, later underpinning GPT, BERT, LLaMA, and contemporary AI advancements. ✨ Personal Insight: This paper underscores the power of simplification, highlighting how groundbreaking ideas emerge by streamlining complexity (e.g., RNNs, CNNs) to focus on essentials. #AI #MachineLearning #DeepLearning #Transformers #AttentionIsAllYouNeed #LearningJourney

17

🤓My husband and I argue a lot about languages. After all , French Hindi English are languages that give you so many pros and cons to argue over. 📣But this is about SQL, the preferred language in the DB world and some other classic query languages vs. their applicability in the multimodal AI world. How about a trip down the memory lane at ApertureData and where that led us? Read about: ❓What multimodal AI data queries look like and require from the database ⛔Why existing languages fell short 💡What was our way out of it when building ApertureDB 🤗How do we still keep it simple for our users Blog link in comment below 👇 :

20

4 Comments

Wrote an article on Byte Pair Encoding (subword Tokenization) and how it works which was potentially used in the the earlier versions of GPT / traditional Language Learning model paradigms. Took a sample of Hindi Language corpus text data to showcase the same. Article below - https://lnkd.in/gCu7Bgpf

33