Harsh Tambi

Fractal, a Mumbai-based AI firm, launched Fathom-R1-14B, a 14B-parameter open-source language model excelling in math reasoning. It outperforms o1-mini and o3-mini, nearing o4-mini levels, costing $499 post-training. Available on Hugging Face and GitHub under MIT license, it was built under IndiaAI’s mission and scores up to 76.7% and 56.7% on AIME-25 and HMMT-25 exams with extra compute. Fine-tuned with supervised learning and model merging, a variant using reinforcement learning costs $967. Fractal also created Vaidya.ai, a free healthcare AI platform, while Sarvam launched Sarvam-M, a 24B-parameter model under IndiaAI.

401

6 Comments

Recently, PW (Physics Wallah) launched Aryabhata 1.0, an AI model built using something called an SLM, which stands for Small Language Model. Unlike big models like GPT-4 or Gemini (called LLMs – Large Language Models) that need a lot of data, power, and memory to run, SLMs are smaller, faster, and easier to use. They are made for specific tasks and give quick answers without using too many resources. This makes them a good fit for education, where speed and accuracy are more important than handling every type of question. While LLMs try to do everything, SLMs are focused — perfect for things like doubt-solving, tutoring, and exam help. Note: A lot of news has surfaced claiming that it is just a wrapper on top of the Qwen model. However,this post is solely focused on how LLM and SLM models differ.

730

8 Comments

🚀 Initiative to build an Open-Source LLM for Konkani Konkani, an official language of India, has a rich heritage, diverse dialects, and it is written in multiple scripts across different regions: ✅ Devanagari – Officially recognized in Goa, widely used in literature and formal settings. ✅ Roman (Romi Konkani) – A legacy of Portuguese rule, still dominant in media, Tiatr, and religious writings. ✅ Kannada – Used in Karnataka, particularly in Uttara Kannada. ✅ Malayalam (Obsolete) – Historically used in Kerala, but most Konkani speakers there now use Devanagari. ✅ Perso-Arabic (Obsolete) – Once used by Konkani-speaking Muslims, incorporating Urdu & Arabic influences. While Devanagari holds official status, limiting AI models to a single script ignores a significant portion of Konkani speakers. Many native speakers read, write, and communicate in different scripts, making it crucial to incorporate all of them when building effective AI models. Konkani’s digital presence is extremely limited, and most existing datasets are proprietary. By leveraging all available scripts, we can enhance AI models, even if the scripts differ, their underlying linguistic structure is closely connected. A script-inclusive approach strengthens the model’s ability to generalize across variations, and should benefit different NLP applications. 💡 The Initiative Launching an open-source effort to build a high-quality Konkani corpus across all scripts; no restrictions, no exclusions. 📌 The first version of the dataset, extracted from Wikipedia, is now available on Hugging Face. However, this is just the beginning! The current dataset is very limited, and we need a much larger, high-quality corpus to build strong AI models for Konkani. 📢 Call to Action If you’re a researcher, student, media outlet, or language authority, we invite you to contribute data, resources, and expertise. Without a strong digital presence, Konkani risks fading in the AI era. Let’s change that together. Codebase & Dataset Repo GitHub: https://lnkd.in/dw4gwBu2 Hugging Face: https://lnkd.in/daaKgxDG This initiative requires collaboration from various stakeholders—researchers, government authorities, academicians, language experts, and tech enthusiasts. Let’s build a state-of-the-art Konkani LLM together! #Konkani #AI #LLM #NLP #OpenSource #HuggingFace #MachineLearning #ArtificialIntelligence #GenerativeAI Government of Goa Goa Konkani Akademi Goa University World Konkani Centre

107

18 Comments

Ever wondered what the IIT Madras BS Programs are actually like?? Reality Check from the campus to the curriculum, from the structure to the experience? The wait is over. The latest episode of the Mathematically Inclined Podcast by Neha Agrawal dives deep into the realities of the BS in Data Science and the BS in Electronic Systems—filmed live from the IIT Madras campus. If you’re curious about the vision, structure, and impact of these unique degrees, this is your behind-the-scenes pass. 🎧 Full episode out now on YouTube: Link in the comments below #IITMadras #engineering #DataScience #ElectronicSystems #iit #jee #MathematicallyInclined #NehaAgrawal #HigherEducation #OnlineLearning #IITMadras #EducationInnovation

185

1 Comment

Proud to Announce: BharatGen - Param 1 | India’s Bilingual Foundation AI Model #AIforBharat #IndicAI #DigitalIndia Introducing Param 1, a 2.9-billion parameter bilingual foundation model trained on high-quality, culturally rich datasets in English and Hindi — powering the next generation of Indic AI systems. Developed under the BharatGen initiative by TIH Foundation for IoT & IoE, Param 1 is optimized for text generation and completion, built on approximately 5 trillion tokens of domain-diverse Indian data. It delivers state-of-the-art performance on bilingual tasks while maintaining computational efficiency. Key Highlights: • English-Hindi bilingual foundation • 2.9B parameters • Trained on 5T+ culturally grounded tokens • MIT licensed and openly accessible • Hosted on AIKosh, IndiaAI’s official platform This is a proud step towards India’s vision of sovereign, inclusive, and accessible AI. Explore more and access the model here: https://lnkd.in/dHcBAD66 Developed by: TIH Foundation for IoT & IoE | BharatGen Team #ArtificialIntelligence #LanguageModel #OpenSourceAI #IndiaAI #TechForGood #BharatGPT

40

🎮 India's Gaming Landscape Gets a Major Overhaul: New Online Gaming Bill 2025 India just passed groundbreaking legislation that will reshape the entire gaming ecosystem. Here's what professionals in tech, gaming, and digital entertainment need to know: 🚫 What's Being Banned All real-money gaming: Fantasy sports (Dream11, My11Circle), rummy, poker, MPL - if you pay to win money potentially, it's out. Complete ecosystem shutdown: No ads, celebrity endorsements, or payment processing for money games. Banking restrictions: Financial institutions are banned from facilitating transactions with these platforms. ✅ What's Getting Support: Esports officially recognized: Government backing for competitive gaming with legitimate prize structures. Social gaming thrives: Casual games like Ludo, Candy Crush, and educational platforms remain unaffected. 💼 Business Impact: The numbers are staggering: ₹25,000 crore industry is facing major disruption. 200K+ jobs are potentially at risk, but this time, I believe the cause is good. 🎯 Let's Imagine The Bigger Picture This isn't just regulation - it's a strategic shift towards protecting consumers from addiction and financial harm. Legitimizing esports as a career path. Creating clearer boundaries between gambling and gaming. For professionals in this space, the companies that adapt quickly to focus on skill-based competitions, educational gaming, and legitimate esports will likely emerge stronger. What are your thoughts on how this will reshape India's digital entertainment landscape? #GamingIndustry #IndiaGaming #Esports #DigitalIndia #TechPolicy #StartupEcosystem

31

1 Comment

Thrilled to be mentoring startups at the UNDP VAYU Hackathon, where innovators are tackling one of the most pressing global challenges—air pollution and sustainable urban mobility. With India’s cities grappling with severe air quality issues, it's imperative to develop scalable, tech-driven solutions that reduce emissions, promote green transport, and drive policy action. VAYU, an initiative by UNDP India, brings together brilliant minds to innovate in: ✅ Emission Monitoring & Data Analytics ✅ Sustainable Transport Solutions ✅ Behavioral Change & Policy Innovation Excited to support the next wave of climate-tech entrepreneurs as they push the boundaries of AI, IoT, and clean mobility to drive real-world change! If you're passionate about climate action, air quality, open networks and sustainable mobility, let’s connect and build solutions that matter! To learn more visit: https://lnkd.in/gQp_iicQ #UNDPVAYU #CleanAir #SustainableMobility #InnovationForImpact #ClimateTech #Mentorship

56

2 Comments

Super excited for Sarvam-M to be opened up to the world! A 24B hybrid model that reads India’s languages, solves maths, and writes code with ease. 🌟 Why we’re excited - Made in India, for India – average +20 % uplift on Indic-language benchmarks, with a massive +86 % jump on romanised GSM-8K. - Brains and speed – a single model with switchable “think / non-think” modes for chain-of-thought depth or fast replies. - Punches above its weight – outperforms Llama-4 Scout and rivals 70 B-class models like Llama-3.3 on most tasks. - Math & code wizardry – +21.6 % on maths, +17.6 % on programming; HumanEval 88 % pass rate! - Lightning inference – FP8 post-training quantization + look-ahead decoding deliver up to 2× throughput on H100 GPUs. Sarvam is building an end-to-end AI ecosystem for India, empowering consumers, businesses, government bodies, and universities to plug cutting-edge models straight into their workflows. Get started with the open weights on HuggingFace. Detailed technical blog: https://lnkd.in/gpCwWSgJ #SarvamM #OpenSource #SovereignAI #IndicAI #LLM

239

5 Comments

The next AI superpower won't be the one with the most GPUs. It’ll be the one that speaks to a billion people in their own language. India isn’t just building another LLM. It’s building BharatGen—a multilingual, multimodal AI rooted in our context. And with Sarvam AI leading the IndiaAI Mission, we're not racing to catch up. We’re carving a new path: → AI that understands regional nuance → Infrastructure that supports sovereign innovation → Leadership that bets on inclusive, homegrown intelligence I see this as more than a tech milestone. It’s a shift in mindset: from importing intelligence to owning it. India’s future in AI won’t be decided in Silicon Valley. It’ll be shaped in places like Hyderabad, Bengaluru, and IIT labs across the country. If you're a tech leader building for India—this is your moment to think local, scale global. The world has built AI for the West. Now, we build it for the rest. #AI #IndiaAI #Leadership #DigitalSovereignty #BharatGen #VishalRustagi #TechForIndia

59

7 Comments

Making sense of that Attention (Q, K, V) formula I recently conducted a detailed lecture on Self-Attention with Trainable Weights as part of Vizuara’s Transformers for Vision Live Bootcamp. In this session, we went line by line through one of the most iconic equations in deep learning - the Attention(Q, K, V) formulation from “Attention is All You Need”. This equation initially feels abstract, but when you break it down visually, you begin to see that it is nothing but a beautifully structured transformation process: 1) We start by transforming the input X into three different spaces - Queries (Q), Keys (K), and Values (V) using separate weight matrices. 2) We then take the dot product of Q with K to measure similarity between tokens. 3) We scale it by √dk to maintain a stable variance. 4) We apply softmax to convert similarity scores into probabilities, giving us the attention weights. 5) Finally, these weights are used to take a weighted sum of the Values (V), giving us the context vector for each token. This visualization you see here was used in my lecture to help learners “read” this formula intuitively rather than just memorizing it. Once you see how Q, K, and V flow through the model, the math starts to make perfect sense. At Vizuara, we believe in teaching deep learning concepts visually and intuitively, so learners not only know what happens but also deeply understand why it happens. If you are someone who has always wanted to truly understand how Transformers work at a mathematical and architectural level, the ongoing Transformers for Vision Live Bootcamp might be the perfect place to start.

1,639

16 Comments

There is big chatter about India developing an AI model. At the end of the day, most of the AI models essentially do multivariate regression analysis utilizing the massive computing power. The objective of any regression analysis is to minimize the loss function, and that is where Deepseek showed jugaad so as to say... Most data scientists say 80% of their time is spent on data-related tasks and only 20% on model development. There is where India should focus on - getting its data right... Otherwise its garbage in and garbage out.

124

33 Comments

Fathom-R1-14B, developed by the AI Research team at Fractal, was featured in Mint in a recent roundup of India's open-source AI models - recognized as one of the most downloaded among homegrown releases in May 2025.   Fathom is a powerful 14B parameter reasoning model designed for accuracy, transparency, and adaptability. It can also support a wide range of enterprise-grade applications, while enabling researchers and developers to build responsibly on top of it.   Still early days, but encouraging to see momentum building for open, reasoning-focused models developed in India.   Model: https://lnkd.in/dF6Qxs-d

179

3 Comments

Scaler x IIT Roorkee is here to redefine your AI learning journey. We’re launching the Advanced AI & Machine Learning Program in association with CEC IIT Roorkee, designed to help you master GenAI and Machine Learning skills to make you ready for the fast-evolving AI world. Here’s what you get: 🔵 IIT Roorkee Certification 🔵 GenAI, ML & AI Engineering Skills built for today’s job market 🔵 Expert Instructors from both Scaler and IIT-R 🔵 2-Day Campus Immersion experience at IIT Roorkee 🔵 Job-Ready Portfolio that recruiters actually notice Anyone opting for the DSML or another Scaler program also stands a chance to be eligible for this module. Check out the program in detail here: https://lnkd.in/djAN6fic #Scaler #IITRoorkee #GenAI #AIEngineering #MachineLearning #TechCareers Anshuman Singh | Abhimanyu Saxena | Amar Srivastava | Naman Bhalla | Rahul Karthikeyan

34

We're excited to announce that we have partnered with bequant.dev ! Through this collaboration, BITS Pilani, Pilani Campus students will get access to : • Curated prep tracks for Quant Development, Research, and Trading roles • Real interview-style problems and learning analytics • Hands-on competitions and discussion rooms led by practicing quants This partnership will enable our students to prepare more effectively while gaining valuable exposure to the standards and expectations of leading global trading firms. The quant culture at BITS Pilani, Pilani Campus continues to thrive, with several leading global trading firms joining us each year for recruitment. Through this partnership, we aim to further strengthen our industry connections, welcome more top-tier firms in the coming years, and ensure our students are well-prepared to meet — and exceed — the expectations of global trading firms. A big thank you to Rajat Jain and Krishnam Gupta, distinguished BITS Pilani alumni, for making this partnership possible and for their continued support. We look forward to many more exciting collaborations with bequant.dev ! #BITSPilani #quant #quantfinance #quanttrading #quantresearch #bequant #finance #python #Cplusplus #careers

75