Aditya Trivedi

"How deep should I go in ML?" This question has always bothered me, where there is so much to learn. -New models emerge weekly. -LinkedIn is flooded with AI breakthroughs. -Research papers are published at an insane pace. So as an ML enthusiast, what should I do? Should I dive deep into every new model? Is it worth it? What if it becomes obsolete next month? A couple of weeks ago the entire LinkedIn was about DeepSeek. It is impossible to keep up with everything deeply. I hate doing superficial work, but I can’t master every new development. That is why I have decided to eliminate noise. I want to focus on what has long-term value, not short-term hype (not saying DeepSeek is noise - it is fascinating). But distinguishing signal from noise is hard. The one thing I know for sure: is that foundational knowledge is invaluable. It builds confidence. It is timeless. But even foundational knowledge is an endless ocean. I could spend: -A month on Lasso regularization. -Three weeks on bias-variance trade-off. -Eight months on Graph Neural Networks. How much is too much? What’s the right depth? I developed a technique to answer this: I ask myself, “Do I truly understand this, or am I fooling myself?” It is easy to believe we know something- until we explain it. So here is my method (since the last few months): I record a quick lecture on an interesting topic. As I speak, I pinpoint my weak areas. If I realize gaps, I trash the recording, learn more, and try again. I repeat until I am confident. This has been the most effective way I have learned. I retain concepts longer, and the clarity I gain is unmatched. The downside? Teaching takes time- time I could spend building AI products at Vizuara. But the satisfaction of simplifying and sharing knowledge is worth it. Am I saying ignore new AI trends? No. Stay aware, but don’t get distracted by hype or FOMO. Have a vision, stick to it. There are people who know a lot of information- the latest news, models, gadgets-but nothing deeply. Don’t confuse information with knowledge. ***** If you wish to learn ML and LLMs from scratch check this out: https://lnkd.in/gua3g9cn

134

4 Comments

Want to ace your next AI/ML interview? Most engineers know about embeddings, but not the full vocabulary of vector databases that power RAG and search systems. This guide breaks it down clearly: ▪️Vector embeddings & similarity search ▪️ANN vs. exact search (speed vs. accuracy) ▪️Indexing methods: Flat, HNSW, IVF, PQ, IVFPQ ▪️Key performance metrics: recall, latency, throughput, QPS ▪️Scaling concepts: filtering, sharding, reindexing ▪️Popular DBs: FAISS, Pinecone, Milvus, Weaviate, Qdrant, Chroma 👉If you’re serious about RAG, search, or AI infra, you need to know these terms. ➕ Follow Naresh Edagotti for more content that makes complex AI topics feel simple.

153

36 Comments

Mythology meets machine learning: India's first case study in AI-powered storytelling at scale. There's been legitimate skepticism about AI in entertainment—can technology truly capture the emotional weight of something as sacred as the Mahabharat? What Collective Artist Network & JioHotstar have executed with Mahabharat: Ek Dharmayudh offers a compelling answer. This is production infrastructure, not just production assistance: ✅ 100 episodes with cinema-grade VFX ✅AI-driven character modeling for expressive consistency across the series ✅Workflows that compress multi-year timelines into months ✅Visual ambition that would've been cost-prohibitive a decade ago Today's AI pipelines enable higher visual standards (as compared to the ones we have grown up with) within scalable budgets—turning what was once a financial barrier into competitive advantage. This marks India's evolution from AI as a tool → AI as core infrastructure for entertainment. 💡The real question: If AI can industrialize the Mahabharat while maintaining cultural depth, what's the next frontier? Original Indian epics designed for global audiences? I would suggest to embrace rather than fighting it, change can be scary but it the only constant #AIInnovation #CulturalTech #ThoughtLeadership #StreamingStrategy #Mahabharat #AI

95

20 Comments

Unlocking the Full Potential of Language Models through Fine-Tuning In the rapidly evolving world of AI, pre-trained Large Language Models (LLMs) have revolutionized natural language processing—but their true power emerges through fine-tuning. I recently came across a concise yet powerful guide authored by Himanshu Sharma that elegantly outlines how fine-tuning transforms a generic model into a domain-specific expert. Think of it like training a skilled novelist to write precise technical manuals—same foundation, refined expertise. Here are a few key takeaways that stood out: • Fine-tuning bridges the gap between general knowledge and specialized performance. • Techniques like Parameter-Efficient Fine-Tuning (PEFT) and Instruction Tuning are making advanced LLMs accessible even to teams with limited computational resources. • Few-Shot Learning offers a flexible alternative when labeled data is scarce. • Domain-Specific Fine-Tuning is critical for industries like healthcare, law, and finance where context matters. Choosing the right fine-tuning approach depends on: • Available compute • Dataset size • Task complexity • Required precision As organizations increasingly look to LLMs for production-ready solutions, understanding how to fine-tune effectively is no longer optional—it’s essential. If you’re exploring AI or building domain-specific tools with LLMs, I highly recommend diving into this material. Feel free to reach out if you’d like to discuss strategies or implementation approaches. #AI #LLM #FineTuning #MachineLearning #NaturalLanguageProcessing #PEFT #FewShotLearning #TechLeadership #InstructionTuning #DomainSpecificAI

50

📺 All in One - The 90% of AI and Data Science on Single Channel CampusX by Sir Nitish Singh Here are everything you need to learn from Python to MCP! 1. 100 Days of Python Programming https://lnkd.in/dwMu-Dtf 2. 100 Days of Machine Learning https://lnkd.in/d4YzcAkY 3. 100 Days of Deep Learning https://lnkd.in/d7Rs_CKz 4. Machine Learning Projects https://lnkd.in/d4MsUEKa 5. Practical Deep Learning using PyTorch https://lnkd.in/dAW27kUA 6. Generative AI using LangChain https://lnkd.in/d442x9GC 7. Agentic AI using LangGraph https://lnkd.in/dt2U-phb 8. Model Context Protocol https://lnkd.in/dUKCHz23 9. FastAPI for Machine Learning https://lnkd.in/dVBhwwSj FOLLOW Umer Haddii for more simplified AI resources! #ai #datascience #machinelearning #deeplearning #genai #agenticai #mcp

33

7 Comments

***** Algorithm of Thought (AoT) ***** Most of us have tried clever prompts to get better answers from large language models. But there is a new way of thinking called Algorithm of Thought (AoT). Instead of only giving instructions and hoping the model figures it out, AoT teaches the model how to explore a problem space. It does not settle for the first answer; it searches, tests, and refines, similar to how a person reasons through a tough question. Why this matters: More accurate and reliable results Handles complex, multi-step problems better Makes LLMs feel smarter and more useful in real work This feels like the shift from guessing to thinking. The prompting may not be about writing the perfect question but about guiding the model’s reasoning itself. #AI #PromptEngineering #LLM #AlgorithmOfThought #MachineLearning

28

8 Comments

In December 2024, Google Researchers published a paper titled "TITANS: Learning to Memorize at Test Time" Siddhant Rai wrote a very detailed blog post on Vizuara Substack regarding this paper. You can read it here: https://lnkd.in/gWAKcdkM I believe this study holds immense potential for LLMs. Here is why: 1️⃣ Overcoming Transformer’s Quadratic Scaling Issue: Traditional Transformers struggle with long sequences due to quadratic time complexity in the attention mechanism. Titans address this by implementing a more efficient memory mechanism that scales linearly with input length, allowing for the processing of much longer sequences efficiently. 2️⃣ Inspired by Human Memory Systems: The Titans model is inspired by how memory functions in the human brain. Memory is fundamental to human cognition, enabling learning, adaptability, and decision-making. Without it, both humans and animals would be limited to basic reflexes and repetitive behaviors. 3️⃣ Long-Term Memory at Test Time: Unlike conventional models that discard past inputs after inference, Titans can memorize information during test time. This allows the model to store and recall knowledge persistently, improving adaptability and making AI behavior more dynamic over time. 4️⃣ Beyond Standard Attention Mechanisms: Transformers rely on self-attention to process entire input sequences at once, but this comes at a computational cost. Titans introduce a different approach to memory that enables efficient knowledge retention without the same quadratic scaling problem. 5️⃣ Potential for Infinite Context Windows: With its scalable memory approach, Titans theoretically allow for infinitely long context windows, meaning the model can retain and utilize information across an extended sequence without forgetting past inputs—something Transformers struggle with due to their fixed context limitations. Here is the blog post by Siddhant for a more detailed understanding: https://lnkd.in/gWAKcdkM Thanks for the incredible post Siddhant! Pic credits: Original Google paper: https://lnkd.in/gBzrP7fA

95

3 Comments

Glad to see that Indian met office is using AI models to predict monsoonal rains. They used Numerical weather prediction models earlier, which had computational challenges and would not work too accurately at fine-grained geographical area. They have built on top of two models, one is developed by European Centre for Medium-Range Weather Forecasts (ECMWF) and the other by Google, known as NeuralGCM. We were contributing and applying SpectFormer (our spectral transformer) for similar AI based weather forecasting instead of using NWP in a PoC collaborating with MSR and met office in a European country about 3 years ago - so quite aware of ForecastNet and other foundational models in this space. https://lnkd.in/gqgXZk5v

18

1 Comment

🚀 Accelerating Responsible AI in India: IIT Bombay Releases 16 Open Datasets on AIKOSH 🇮🇳 . Indian Institute of Technology, Bombay Bombay has taken a significant step toward advancing responsible and inclusive AI by contributing 16 diverse datasets to AIKOSH, the Government of India’s official AI repository. These open datasets reflect India’s cultural and linguistic richness and are aimed at fueling research and innovation in Artificial Intelligence (AI) and Machine Learning (ML) — tailored for Indian use cases. 🔍 What’s Inside: Handwritten & printed Indian scripts Tables extracted from scanned documents Multi-language audio samples Drone surveillance imagery Visual question-answer datasets 🎯 Why It Matters: These datasets help AI models: Recognize and interpret Indian languages and handwriting Process documents and media with regional nuances Analyze audio-visual data in Indian contexts This initiative marks a crucial step toward building AI that works for India, by India — empowering researchers, developers, and institutions nationwide. 🔗 Explore the datasets on AIKOSH: https://lnkd.in/gJv4XS4j #AI #MachineLearning #ResponsibleAI #OpenData #IITBombay #AIKOSH #IndiaAI #Innovation #Research #DigitalIndia

31

5 Comments

50% of my college time was spent learning ML. I specifically focused on generative models and computer vision. Even after college, I worked for 2 years in computer vision. I had very little experience with NLP. But then it took me just 6 months to cover classical NLP, NMT,and attention, and start working with language models like BERT, GPT etc. And my learning philosophy remained the same. Learn in-depth maths, write code and explore libraries. This speed-up was almost like transfer learning for me. This is true not only for ML but also for other fields. Be it backend, frontend, or devops. As you learn more, you’re able to connect the dots and fill in information gaps faster. You don’t need to acquire all the information to understand a concept since things are more or less the same or follow a similar framework.

544

13 Comments