Pranali Yawalkar

We had a great session about NVIDIA NIMs with Anshul Jindal and Mark Moyou, PhD yesterday. 🚀 From Mark: We got lots of info on inferencing with NIMs and different types of NIMs like embedding models, generative models, reranking models.. 🎨 I walked through how you can use NIMs for embedding, retrieval and generation with #Haystack 🚀 And finally, Anshul did a live coding session where he showed how you can deploy and scale your own self-hosted NIMs. If you missed it, or don't even know what I'm talking about but you wanna learn, the recording is up (adding to the comments👇) All resources, including the example repos and guides are in the description. #AI #LLM #Nvidia #opensource

36

2 Comments

🚀 New Blog Post Alert: NormCompressAI – Exploring Image Compression with Different Norms! 🎨📉 In the world of deep learning and computer vision, image compression plays a crucial role in optimizing storage and transmission efficiency without sacrificing quality. My latest blog delves into NormCompressAI, a project utilizing autoencoders to compress images and examining how various mathematical norms (L1, L2, L-infinity) impact compression quality. 🌟 What's Inside: - Detailed exploration of image compression using a custom autoencoder architecture - Experimentation with different norms to assess image reconstruction quality - Utilization of the CIFAR-10 dataset as a testing ground - Implementation in PyTorch, inclusive of code and results - Visualizations and loss curves to showcase the performance of each norm Curious about AI and image compression convergence or keen on norm experimentation? Check out the full blog post here: https://lnkd.in/g3znK55r 🔗 Full project code accessible on GitHub: https://lnkd.in/grV_Zp6c Feel free to share your thoughts and don't hesitate to tinker with the code for varied datasets or adjustments! 😎 #AI #DeepLearning #ImageCompression #Autoencoders #PyTorch #MachineLearning #ComputerVision #DataScience #Norm

15

Thanks to the DeepLearning.AI team for this in-depth course on Finetuning LLMs.

1

This tutorial by Vipula Rawte and Aman Chadha aims to bridge this knowledge gap by introducing the emerging realm of hallucination in LLMs. It comprehensively explores the key aspects of hallucination, including benchmarking, detection, and mitigation techniques.

35

2 Comments

The focus alternatively can be to concentrate on more niche, yet fundamental, challenges in AI, honing LLMs to perform specific tasks, tackling the data challenges inherent in LLMs, and improving transfer learning from English to other languages. There is also critical need for efficiency and the integration of external systems and application through agentic frameworks .. Read more at: https://lnkd.in/gzy7naZz

11

🚀 Built a Transformer Model from Scratch! 🚀 I recently delved into constructing a Transformer Model from the ground up, embracing the transformative power of natural language processing. Here are 8 key takeaways from this enlightening journey: 🔹 Tokenization Matters: Effective tokenization is vital for meaningful sequence modeling, exploring methods like WordLevel and Whitespace. 🔹 Greedy Decoding: Implemented a step-by-step greedy decode algorithm for precise sequence generation. 🔹 Masking for Causality: Utilized causal masks to maintain the autoregressive nature, focusing on past tokens. 🔹 Cross-Attention Mechanism: Explored the intricate connection between source and target sequences through cross-attention mechanisms. 🔹 Handling Variable Sentence Lengths: Dynamically managed varying sentence lengths by determining optimal source/target lengths. 🔹 Loss Function Design: Enhanced training stability and generalization with label smoothing and padding token exclusion. 🔹 Dynamic Learning Rates: Employed learning rate scheduling to fine-tune optimization and prevent overshooting. 🔹 Model Checkpoints: Implemented model checkpoints to preserve training progress and prevent data loss. Witness the remarkable progress of my model, elevating from initial predictions to surprising translations in just one epoch of training. Explore the project here: [GitHub - Transformer-based-Translation-Model](https://lnkd.in/dQd5ecwC) If anyone wants to pair on exciting ML projects feel free to connect and message me 😊!

35

6 Comments

In GPUStack 0.2.0, we added :multi-GPU inference, distributed inference across workers, CPU inference, and binpack & spread strategies. Whether you need to assign specific workers or manually select GPUs, GPUStack 0.2 provides the flexibility and performance for handling advanced LLM workloads. Please download to trial it, any your feedback are so appreciate!

2

3 Comments

Bhaskar Lalwani and I (Aniruddha Mukherjee) had worked to create kabaddiPy recently. kabaddiPy is an open-source repository for Kabaddi data - a first step towards making this fast-growing, dynamic sport more accessible for fans, analysts, and researchers. Kabaddi is the second most-watched sport in India after cricket, yet its data is not as widely available. We hope to be the trailblazers and change this! Whether you're into sports analytics 📊, passionate about Kabaddi 🤼‍♂️, or simply love data, we invite you to: ⭐ *Star the repo* 🔧 Contribute to the data 💡 Explore insights (by building predictive ML models / more!) Open an issue and let's chat! Let’s work together to build something great for India 🇮🇳 and the Kabaddi community! Link to the repo: https://lnkd.in/gGCnY_B4 Link to docs (for an overview): https://lnkd.in/g-Khu5_J btw, we're a part of GSSoC Extended and Hacktober-Fest as well! #opensource #kabaddi #sportsanalytics #dataanalysis #ProKabaddi

79

1 Comment

GATE 2025 offers the flexibility of choosing two paper combinations. Exciting news from IIT Roorkee as they unveil the tentative list of GATE 2025 Paper combinations! If Data Science & AI is your primary focus, but you also have an interest in Computer Science, ECE, or other branches, the best GATE Paper combination awaits you.🌉🚀 🥴 Check out the snapshot shared by EECS Academy from the official GATE 2025 website. #GATE2025 #IITRoorkee #DataScience #AI #ComputerScience #ECE

9

1 Comment

🚀 Excited to share some early progress on my **Image Search using Natural Language Queries** project! This work is inspired by OpenAI's CLIP (Contrastive Language-Image Pre-Training) and aims to create a more intuitive and powerful way to search for images using natural language descriptions. 🛠️ I'm still in the early stages, and while the results aren't perfect yet, it's a promising start. If you're curious about the technical details or want to see the code in action, you can check out my Jupyter Notebook here https://lnkd.in/eUDZr8yy 🔍 The journey has just begun, and I'm looking forward to refining the model further and exploring new possibilities. #AI #MachineLearning #NLP #ComputerVision #DeepLearning #Innovation #ImageSearch

28

1 Comment

𝗖𝗵𝗲𝗺𝗖𝗵𝗮𝘁 | 𝗟𝗟𝗠-𝗣𝗼𝘄𝗲𝗿𝗲𝗱 𝗖𝗼𝗻𝘃𝗲𝗿𝘀𝗮𝘁𝗶𝗼𝗻𝗮𝗹 𝗔𝘀𝘀𝗶𝘀𝘁𝗮𝗻𝘁 𝗳𝗼𝗿 𝗠𝗮𝘁𝗲𝗿𝗶𝗮𝗹 𝗦𝗰𝗶𝗲𝗻𝗰𝗲 𝗮𝗻𝗱 𝗗𝗮𝘁𝗮 𝗩𝗶𝘀𝘂𝗮𝗹𝗶𝘇𝗮𝘁𝗶𝗼𝗻 I am pleased to share that our IBM Research project 'ChemChat' received another very enthusiastic response at the MRS Spring Meeting 2024 in Seattle - 6 weeks after an engaging talk at the symposium on "Exploring the Impact of ChatGPT and Large Language Models on Cheminformatics" during the ACS Spring 2024 in New Orleans. This marks the fourth in-person presentation of our ever advancing proof-of-concept prototype application since end of February 2024 and, this time, we transformed our poster session into an expo-like experience, demonstrating the capabilities of 'ChemChat' live. Attendees participated in hands-on demos, creating a continuous flow of interest and a steady crowd throughout the evening session. The engagement and feedback from academia and industry were outstanding and we are very grateful for the constructive discussions and advise. Our team is proud of the progress made since the inception of the project in February 2023. We are now focused on implementing new features in collaboration with colleagues across various IBM Research labs to unlock the synergy behind using internal and external expert tools in conjunction. Furthermore, we are exploring ways how to seamlessly onboard new services and how to process complex, chained queries. Our biggest goal for this year is to advance 'ChemChat' to a stage where we can begin trials with selected partners. The request on how to access 'ChemChat' is always the very first question during our live demos - and so is the waiting list growing. #ibm #ibmresearch #chemchat #acsspring2024 #mrsspring2024

74

7 Comments

I remember when this all started. Faris - invited me to catch up at our favorite spot around Blok M. I had no idea he’d just wrapped up a procedure at Pondok Indah Hospital! (He really knows how to add drama to our catch-ups.) We’d been on separate paths and projects since "the winding down project", but there he was, reminding me of our dream to ride this AI wave together again. It was a plan we’d hatched with Sahal Zain & Gurnoor Dhillon when the world of AI was still mostly theoretical and jargon-heavy. Somehow, the conversation continued with Gurnoor Dhillon while he was waiting for his meeting with someone who would eventually become one of CalvinBall’s key early adopters. It felt like everything was coming together, like a series of seemingly random dots connecting to form a bigger picture. Seeing how far we’ve come as a team at CalvinBall Technologies is truly exciting. What started as theoretical discussions has now evolved into practical Gen AI solutions that are solving everyday problems and hold endless possibilities to shape the industry. 👨‍🚀 👩‍🚀 Here’s to pushing boundaries and achieving new milestones together! Gurnoor Dhillon, Sandeep Ramesh, Faris -, Sahal Zain, Eva Zuliya, Ratu Nadiah Khairunnisa 🚀 #GenAI #AIInnovation #TechJourney #ArtificialIntelligence #TechTransformation #CalvinballTech #Calvinballers

55

2 Comments

In just about a month since being made available as a pre-print, my research paper titled 'Deep Learning for Climate Downscaling: Generating high-resolution gridded temperature projections over India from low-resolution CMIP6 data' has reached the milestone of 100 reads on ResearchGate, which is a professional network for scientists and researchers to share papers, ask and answer questions, and find collaborators. The paper has also received around 300 reads on ESS Open Archive, a community server established to share early research outputs in earth, environmental, and space science. The paper has been read by researchers from various countries, including India, UK, US, Spain, Brazil, Jordan, China, the Netherlands, Germany, Japan, Australia, New Zealand, and the Republic of Ireland. Link to paper: https://lnkd.in/gHtmDwGV

42

1 Comment

Reflections two weeks late are better than never! It was my pleasure to participate at the Shaping AI Futures event hosted by Aapti Institute on 30th September, as an invited speaker on a panel exploring the evolution of Responsible AI in India with Sowmya Karun and Sivaramakrishnan R Guruvayur Ph.D, ably moderated by Soujanya Sridharan. I spoke about ensuring responsible development of AI in the presence of incredible excitement for the technology’s transformative potential for the public sector, emphasizing critical analysis and implementation across five stages of the solution journey: problem selection/definition, data collection, model creation, evaluation and deployment, while contextualising these for Indian realities. I also shared my thoughts on how policy makers and civil society organisations can aid AI development in a manner that protects users and workers but also encourages innovation, emphasising the need to cut through the hype, reject pseudoscientific priors, incorporate technical and economic realities and protect from harms. As part of the industry showcase, I walked the audience through our operating model and solutions portfolio across healthcare, agriculture and education. I also described the Wadhwani AI Responsible AI playbook (https://lnkd.in/dv93ncQH) we wrote after a year of intense research and consultation, demonstrating application of its principles in three different solutions. I enjoyed Sundaraparipurnan Narayanan's keynote, a thought-provoking exploration of the political economy of (responsible) AI and applied AI ethics in varied contexts, conveyed very effectively through a series of cartoons. Another highlight was the roundtable discussing Aapti's new RAI framework, which saw some excellent perspectives and conversations. I'd like to thank the Aapti team for the wonderful event and their warm hospitality: Rattanmeek Kaur, Gautam Misra, Supratik Mitra, Soujanya Sridharan, Vaishnavi Patil and Sarayu Natarajan. It was a privilege to interact, learn from and bounce ideas with experts from very diverse worlds, and I look forward to continuing to do so. #ResponsibleAI #AIEthics #ResponsibleAI #EthicsInAI #AIForGood #AIPolicy #AI

155

2 Comments

❝ How do you taste your Wine? With the right Glass, served from the right Bottle. How do you utilize your LLM? With the right Task, served from the right Compute environment. ❞ If the quote above caught your interest, check out my blog post that was featured on the Toyota Connected India Medium channel last week: Leveraging LLMs in Cloud - 🔗 https://lnkd.in/gfHsdevH --- I hope this blog adds a bit of amusement for the seasoned AI/ML engineers 🧠👨‍💻 with the carefully crafted images and perhaps even the silly wine analogies 😉. Whether you're a cloud engineer ☁️ 👨‍💻, a product owner of cloud-based apps, or someone new to Generative AI in the cloud, this blog offers a detailed, beginner-friendly look into the selection and deployment of LLMs in cloud environments. --- A heartfelt thanks 🙏 to my colleagues at Toyota Connected India for taking the time to review, shape this piece and even share a creative writeup like below in the original post 🫡 👇 🤔 Curious to dive in? Click the link below to explore the blog on Medium! 🌐 🔗 https://lnkd.in/gfHsdevH

18

Did you know that scientists have accidentally discovered a new formula for Pi? 😳 Saha and Sinha who work in the Centre For High Energy Physics, IISc at the Indian Institute of Science (IISc) published a paper on 'Field Theory Expansions of String Theory Amplitudes'. It's in this paper that they mention this new formula for Pi but... It's mentioned in the appendix 🤔 Both scientists were working on string theory amplitudes and it just so happens that while they were working on this research, they uncovered this new formula for pi. What's even more fascinating is that this formula has a constant parameter, lamda. This means they not only discovered one formula for pi but infinitely many 😳 For more information, check out my latest YouTube video👇 https://lnkd.in/eYpWd6dg

37

1 Comment

Update on our FYP: Early Retirement: A Game with Live NPC Interaction Using LLMs Continuing the course of Generative AI from Oracle, I added on to work with Langchain. I've heard a lot about Sir Krish Naik 's course on it, and oh man, the content it covers is really orgasmic😵. Now working on creating my first chatbot, it will take some time, but yeah, will share some snippets once completed. Let's check what Muhammad Asif has done for today, and till then, keep connecting to ride along this FYP journey. #FYP #AI #LearningJourney #EarlyRetirement #FinalYearProject #LLM #Support #GameDevelopment

27

From direct query to increasing sentence context window and complementing it with auto merging increase the performance and quality of the RAG inference. The suit from TruLens are really good to grasp the LLM output. The RAG triad is here. https://lnkd.in/ga63kg7Z

22

3 Comments

I'm excited to share my recent blog post on CEVA's Expert Blog section! 🚀 In this article, I explore the nuances of NPU IP architecture, shaped by software insights and use-case analysis. If you're curious about how software and use-case analysis impact IP architectures, you may find this piece valuable. I welcome your thoughts and insights on this topic. Feel free to share your comments below! #Ceva #NPU #IPArchitecture #SoftwareInsights #UseCaseAnalysis #TinyML

39

4 Comments