Anshum Gupta

If you are building on Hyperliquid, your RPC choice can make or break your performance. We broke down how OnFinality delivers fast, reliable, and scalable RPC for HyperEVM.

Retrieval augmented generation (RAG) has become a popular paradigm for enabling LLMs to access external data and also as a mechanism for grounding to mitigate against hallucinations. RAG models are trained to retrieve relevant documents from a large corpus and then generate a response based on the retrieved documents.#GoogleCloudLearning #GoogleCloudSkillBadge#GenAIExchange

4

Dr.Sheth’s thoughts on building the required talent and ecosystem in India to spearhead its foray into deep tech AI products/companies is both timely and critical, for strategic autonomy over the long term. 

22

3 Comments

Demystifying AI at Stanford Continuing Studies & the HAI AI Index A brief recommendation for two resources: 1. Stanford Continuing Studies - several courses offered online and in person address key demand for AI education for technical leaders, investment leaders and ethicists. https://lnkd.in/gk5MJ2PC. I've attended several rewarding classes here in the past year which expanded my professional network and my knowledge of AI ethics and my ability to track the evolution of the AI industry. In my latest class Demystifying AI, Nestor Maslej treats his class to week-by-week insight at an analytical level not typically available in public news sources. 2. The HAI AI Index - https://lnkd.in/geSqc6Bs - lies at the core of the last course I attended, led by the editor of the AI index (https://lnkd.in/gXu2HkqN). Nestor and the HAI team are looking at the industry in ways that enable meta-viewpoints on progress and capability. I have no affiliation with Stanford - just a grateful review of effective curricula and for gaining the skills needed to peer into & shape our AI future. And sharing an observation on the current era of AI - this is a good time for people who have both technical and legal training to contribute to our AI future.

15

🌐 Aspora Secures $53M to Power Banking for NRIs and Immigrants Aspora, formerly Vance, has closed a $53 million Series B round to accelerate its mission of building modern financial solutions for the Indian diaspora and immigrant communities worldwide. 🔹 The funding was co-led by Sequoia and Greylock, with support from Quantum Light Ventures, Hummingbird Ventures, and Y Combinator. 🔹 Aspora has now raised more than $99 million, reaching a $500 million valuation. 🔹 The platform enables NRIs to send remittances, invest, bank, and access credit and insurance globally. 🔹 In the last six months, Aspora’s transaction volume jumped from $400 million to $2 billion, serving over 250,000 users. 🔹 The company is expanding from the UK, UAE, and EU to the US in July, and to Canada, Australia, and Singapore by year-end. 💡 Why It Matters 🔹 Aspora is bridging the gap for NRIs and immigrants seeking seamless cross-border financial access. 🔹 The company’s growth reflects a broader shift toward digital-first, global banking for diaspora communities. 🔹 Backing from top VCs and angels highlights confidence in Aspora’s vision and execution. 🔭 What’s next? 🔹 Aspora will soon launch bill payments, NRI fixed deposits, and a streamlined banking account. 🔹 Upcoming services will also support healthcare and family needs for users’ relatives in India. #Aspora #Fintech #Banking #Remittance #IndianDiaspora #Startups #Innovation #Sequoia #Greylock #GlobalFinance

9

ACTORS + AI: Director/Actor/Producer/Writer AI Chef 🧑‍🍳 Hans Obma visits the Cozy AI Kitchen 🎂 https://lnkd.in/gjiAvEfP to share his insights on how AI has transformed his creative processes. You may have seen Chef Hans in such 📺 hits as Better Call Saul, Wandavision, or Narcos — and yes, he's probably the most handsome AI 🧑‍🍳 chef we've had in the kitchen to date :-). Hans needed to learn a new language ... how did he do it? He got multimodal AI models to help him speak Welsh. Hans needed a way to scale how he communicated with his stakeholders ... how did he do it? He created an authentic-to-him way of scaling by creating an instant entourage of supporting agents. HT 🧑‍🍳 Ross Heise 🧑‍🍳 Matt Scholz and the power of the Wisconsin network for bringing such a storied talent to the Cozy AI Kitchen! --- This 🍰 episode: https://lnkd.in/gjiAvEfP All 50 episodes: https://lnkd.in/g6upvbGX

89

11 Comments

Google Antigravity exposes the critical flaw in autonomous AI agent architectures: trust based governance at scale. Antigravity represents a major shift in how developers work. Instead of writing code line by line, you delegate entire tasks to autonomous agents that can modify files, run tests, browse the web, and execute changes across your codebase in parallel. The problem? These agents operate on trust, not proof. One developer reported: On Day 3, an agent confidently refactored a utility function and silently deleted a critical edge case check. This is not a bug. This is the inevitable result of autonomous agents operating without cryptographic authority validation. When you have three agents working asynchronously across different files, two critical questions emerge: how do you enforce what each agent is authorized to do, and when multiple agents coordinate on shared resources, how do you maintain isolation between their operations? Policy based guardrails do not work at this scale. I have solved both problems through complementary cryptographic architectures. For individual agent authorization, my 13 layer cryptographic governance system validates AI agent authority mathematically before execution. For multi agent coordination, my YIN COLLAB architecture implements Agent Specific Compliance Tokens with per agent privacy isolation maintaining cryptographic boundaries preventing cross contamination. Every action carries immutable proof of authorization. Every decision boundary is pre validated cryptographically. 26 USPTO patents. 2,330 claims. Validated with 640x timing resistance and 500 plus concurrent agent support with sub 15ms latency. Mathematical proof, not policy promises. For developers using Antigravity, Cursor, or any autonomous agent platform: the question is whether you can prove mathematically that unauthorized operations are impossible and that multi agent coordination maintains isolation. Because as agent orchestration becomes the dominant development paradigm, the liability surface expands exponentially. One misconfigured agent with access to production systems is an organizational failure. One agent leaking sensitive data to another agent through shared context is an architectural vulnerability. Making it mathematically impossible for agents to violate boundaries is the only governance model that scales. Autonomous agents are the future of software development. Cryptographic governance is the only way to make that future safe. https://lnkd.in/ejPREk9D #GoogleAntigravity #AIGovernance #AutonomousAgents #Cybersecurity #AIAgents #DeveloperTools #CryptographicSecurity #ZeroTrust #AICompliance #SoftwareDevelopment #TechInnovation #AIEthics

2

This is a significant partnership for us, the first one with a Big Tech. We look forward to building solutions with Google AI Models (on Vertex AI Platform) to strengthen the account aggregator ecosystem. This will help all participants in the ecosystem, especially the end consumer. / Sashikumar Sreedharan

17

As AI data centers hit power limits, Peak XV backs Indian startup C2i to fix the bottleneck | TechCrunch https://buff.ly/neRBlVr

4

Identity verification shouldn’t take weeks of integrations, custom logic, and stitching APIs from five different standards. Hovi Core SDK gives developers a single, unified way to verify any credential SD-JWT, mDocs, JSON-LD, and more. One API call. One pipeline. Zero complexity. From strong authentication to automated access control and reusable KYC/KYB, you can make your app identity-aware without becoming identity-complex. Build trust once. Scale it everywhere. #Developers #DigitalIdentity #VerifiableCredentials 

2

Mercor, a startup connecting AI labs with domain experts for model training, has raised $350 million at a $10 billion valuation, led by Felicis Ventures V LP with participation from Benchmark , General Catalyst, and Robinhood Ventures. Initially an AI hiring platform, Mercor pivoted to supplying specialized professionals (e.g., scientists, doctors, lawyers) who train foundational AI models, earning the company hourly finder’s fees. It has since expanded into reinforcement learning infrastructure and aims to build a broader AI powered recruiting marketplace. Mercor’s growth accelerated after OpenAI and Google DeepMind reportedly cut ties with Scale AI, following Meta’s $14 billion investment in that company. The firm claims it’s on track to reach $500M ARR faster than Anysphere (Cursor). Currently, Mercor: Pays $1.5M daily to over 30,000 experts earning an average of $85/hour. Plans to focus on: Expanding its expert network Enhancing matching systems Automating internal processes Mercor’s mission is to help AI systems capture economic subtleties, like judgment, intent, and trade offs, that current models often miss. Mercor’s rise signals a powerful new layer in the AI value chain, one that bridges human expertise with machine learning at scale. As foundation models race to master nuanced reasoning and decision making, Mercor’s network of vetted domain experts provides the critical “human feedback infrastructure” these models need to move beyond surface level intelligence. This positions Mercor not just as a staffing platform, but as a strategic enabler of AI evolution, transforming expert insight into structured, monetizable training data. With $500M ARR in sight and deep adoption by top AI labs, Mercor is emerging as a potential “human capital backbone” for the next generation of foundation models, where judgment, ethics, and domain precision become as valuable as compute. #aiinvestment #aiforall #ai

17

AWS today launched its AI Agents SDK/framework. I talked to Neha Goswami, the Director of Developer Agents and Experiences at AWS, about how Amazon's take is different from its competitors. https://lnkd.in/gcX_778y

18

Major developments are on the horizon for India's burgeoning tech and AI ecosystem. Nvidia CEO Jensen Huang is set to visit India next week, engaging with prominent tech leaders and government officials. This visit is strategically timed as India intensifies its focus on domestic semiconductor manufacturing and AI development. Nvidia, a global leader in AI chips, has been a critical enabler for Indian AI startups and data analytics firms. Discussions are expected to revolve around potential investments, R&D collaborations in AI, and bolstering India's semiconductor ambitions. This signifies Nvidia's growing commitment to the Indian market, positioning India as a potential global AI hub. For professionals in AI, semiconductors, and venture capital, this visit could signal new partnership opportunities, talent demand, and a significant acceleration in the local tech landscape. How do you see this impacting India's journey towards becoming a global AI powerhouse? https://lnkd.in/g2h5SiaG #IndiaTech #AIInnovation #Semiconductor

7

Why Guardrails Are Critical for Safe AI Adoption Generative AI has moved from a research novelty to an everyday reality. Tools like ChatGPT, LLaMA and Gemini are redefining how we write, code and create. But with great power comes great responsibility and that’s where guardrails step in. LLMs are brilliant, but they’re not flawless. They can hallucinate facts, amplify bias, leak sensitive data or even generate toxic content. Left unchecked, these risks can lead to misinformation, legal troubles or loss of user trust. So, what are guardrails? Think of them as AI safety layers programmable rules and systems that monitor and control what LLMs produce. They act during user interaction, filtering unsafe inputs and outputs in real time. A good guardrail doesn’t just block bad content; it ensures outputs are accurate, ethical and compliant. Here’s how the ecosystem looks today: ✅ Llama Guard (Meta) – Fine-tuned on the LLaMA architecture to classify content into predefined safety categories. Flexible for different use cases. ✅ NVIDIA NeMo Guardrails – Uses structured conversational flows with KNN-based intent matching and moderation to keep conversations on track. ✅ Guardrails AI – Adds type and format constraints (e.g., enforcing JSON structures), plus corrective prompting if outputs fail checks. ✅ TruLens – Focused on evaluation and feedback loops to improve context relevance, groundedness, and fairness in RAG-based apps. ✅ Guidance AI & LMQL – Programming approaches that combine logic and generation, letting developers enforce constraints using regex, control flows and real-time checks. Why does this matter? Because attackers are getting creative. Jailbreak attacks, prompt injections, and adversarial queries can still bypass basic safeguards. Some even exploit language patterns or use multilingual prompts to trick models into producing harmful content. Building strong guardrails isn’t just about adding filters it’s about system design. We need: ✔ Multi-disciplinary strategies combining neural and symbolic reasoning ✔ Continuous monitoring and evaluation ✔ Integration across the AI development lifecycle, like safety-critical systems in aviation or automotive (think ISO 26262) The ultimate goal is trust. Businesses, governments, and users will only adopt AI at scale if they believe it’s reliable, fair and safe. Investing in guardrails is not an afterthought it’s the backbone of responsible AI. What's Next? Guardrails need to evolve beyond simple filters. The future lies ✔ Neural + Symbolic Systems working together for deep reasoning ✔ Lifecycle integration-from design to deployment (think ISO 26262 for Al) ✔ Continuous monitoring to detect new attack patterns and biases #AI #LLM #ResponsibleAI #Guardrails #GenerativeAI #EthicsInAI #AITrust

1

Ruokai Yin, our summer intern from Yale to Microsoft Azure, worked with Xuan Zuo, @Sattwik Deb Mishra, Hokchhay T. and Preyas Shah and me. We published a paper based on his project: https://lnkd.in/eHuG3BSt Summary of paper: Finding optimal sharding is hard especially for large LLMs. A large number of devices have to be divided into TP, PP, EP dimensions. For TP, we have to decide the tensor dimension e.g. head, hidden size to shard for every op. Additionally, TP can be one-dimensional/two-dimensional etc. This makes for a large sharding space that is hard to cover exhaustively. SoTA methods use heuristics to select sharding e.g. GSPMD, Sequence Parallelism, Helix etc. There are significant time gaps, months or years, between the discovery of each of these methods. We want to fast-forward the process of discovery by 1) using a formal representation of the sharding space, and, 2) using reinforcement learning to search the sharding space.    This paper demonstrates that our method is 3.5x better than random search, and on par with hand-tuned heuristics when looking only at 1D sharding and without allowing sequence parallelism. We aim to extend the work to multi-dimensional sharding as well as sequence parallelism. Beyond that, we wish to combine the problem of sharding with fusion, tiling and collectives as well.

58

We are entering the post-GPU-only phase of AI infrastructure. At Punky Tiger Labs, we believe the next performance and reliability gains in inference won’t come from acceleration alone, but from adding a cognitive coprocessor layer designed for persistent state, orchestration, and continuity. This is the thesis behind NYMPH: not replacing existing compute, but extending it where current stacks still lose efficiency and coherence. If you are an investor, distributor, or hardware partner tracking where AI infrastructure is moving next, this is the conversation to watch. Read the full thesis: PTL Insights: https://lnkd.in/evXRWVVa NYMPH Insights: https://lnkd.in/eC5qcuxR #AIInfrastructure #AIHardware #Inference #DeepTech #Semiconductors #PunkyTigerLabs #NYMPH

2

This is what I’ve been waiting to see out of India. For 25 years, India’s tech story was operational leverage - bodies, not products. The unicorns that followed were American business models localized for Indian scale. Emergent is different, winning on product, not labour arbitrage. The shift is that Indian founders are finally building category-defining global products.

6

🚀 Built to Ship – Hosted by Tyler Postle, Co-Founder of Voker (YC S24) 🎙️ 🎧 Featuring Aizaz Manzar, Chief Agentic AI Product Officer for Sales and Marketing at Amazon Web Services (AWS) In today's episode, I learn why AI isn't failing us—we're just building on shaky foundations. Aizaz shares how 16 years in business roles at Dell taught him that insights without context are meaningless. When he joined AWS 7 years ago, he started building the data infrastructure that would later become the foundation for their agentic AI systems. The problem: Most companies are frustrated with AI because they're using simple chatbot implementations without proper data structures, knowledge graphs, or business context. They expect magic from tech teams alone, but this is a tech + data + business problem. Aizaz's solution at AWS: Build hybrid teams where data engineers understand business, product managers can code, and everyone knows 80% of the business domain. His team owns the full vertical stack—from backend data infrastructure to business context layers to agentic AI products. Key insight: "A good knowledge graph isn't differentiated by technology—it's the business knowledge infused into it that matters."

19

2 Comments