Vinay Rao

Voice is a super interesting modality right now - maybe the first modality we're seeing move to open source models across a number of scale ups / enterprises. Reliability concerns, high costs, and open source model performance are pushing engineers to do their own fine tuning vs. relying on third-party vendors of proprietary models. Many of these orgs have already been collecting their own first-party data and now with third-party vendors like Extrian, David AI, etc they can train really high quality models. RL has been insanely hyped, but it's been unclear how long it will take scale ups and enterprises to actually lean in. Voice AI might be hitting that inflection point faster than expected.

45

7 Comments

Truly transformative. Within the next 6 months coding will be fully commoditized. Only design, distribution, and hardware will be the moat. It is time to lean in #AI #FutureofWork #OhioTech

1

1 Comment

Building on top of large language models is fun, but getting consistent performance and reliability is extremely challenging. I love this post from Kimberly Patron about how Sierra uses health and performance-based traffic routing and request hedging to get meaningfully better performance and tail latency from high scale LLMs https://lnkd.in/g4qWRCb5

284

13 Comments

Stanford just published data that reframes a lot of quiet assumptions about AI. Open weight models are no longer a niche choice. They are becoming a default path for adoption. The charts are telling. Qwen has overtaken Llama in cumulative downloads. DeepSeek is not a one off. Downstream fine tuning and derivative models are increasingly built on open weights. This is not a political story. It is a systems story. Open weights spread faster. Permissive licences compound. Efficiency beats scale. Deployment beats benchmarks. What I find interesting is how little this shows up in most AI strategies. So I am curious. If you are working with AI today, how are you thinking about openness and control. What is driving your choice. #AI #OpenWeight #AIGovernance #ResponsibleAI #AIAdoption #AIStrategy

2

1 Comment

🚀 Just wrapped up a hands-on project applying Retrieval-Augmented Generation (RAG) to a real-world use case: interpreting LIC’s New Jeevan Shanti policy document using LLMs. The core idea? Combine vector search over regulatory PDFs with GPT-4 to generate grounded, faithful answers to complex policy questions. Key capabilities we built: ✅ LCEL-based RAG architecture using LangChain ✅ Contextual compression retriever for focused document lookup ✅ Modular LLM & embedding selection (OpenAI, HuggingFace, Mistral) ✅ Evaluation using LLM-powered metrics: faithfulness, relevance, correctness 📄 The document used is publicly available here: LIC New Jeevan Shanti Policy PDF 💻 Full code available on GitHub: https://lnkd.in/gS9mEzWA 💡 This architecture brings the best of both worlds — retrieval accuracy and generative flexibility — especially in regulated domains like insurance and financial services. 🔜 Next step: Incorporate Agents that can reason across multiple steps — not just answer a question, but decide which tools to use, when to search, how to validate, and when to stop. That’s where the future of enterprise GenAI lies. Would love to hear how others are pushing RAG, agents, and LLMs in production! #GenAI #LangChain #LLM #RAG #AIForOps #InsuranceTech #LLMEvaluation #LangChainAgents

61

1 Comment

Here is a framework for nations to work with revolutions such as #AI and #Quantum. First you have your people learn, then dabble and become early/late adopters (hopefully not laggards!), then become effective technical users and finally value creators. It is only then that the economy benefits from your investment, and you create sustainable job growth. While #India is a key market for #Quantum with millions thirsting to learn and actually do things with this technology, there are still large gaps between: 1. Where Quantum Computing is today Vs where it needs to be for mass adoption [Quantum Sensing on the other hand is ready for prime time and GPS compromising incidents are only increasing] 2. The specialist work being done Vs the ability to hire millions of generalists 3. What it costs to tap into the technology Vs perception of what it should cost as a commodity 4. The talent it takes to produce world class indigenous technology Vs the talent that actually exists on the ground 5... Each of these gaps will close over time. In the lead-up and as a first step, India first needs to learn enough about this technology, ideally from the very best, and spread awareness. The starting point of this journey should ideally not be from dull videos or low quality social media content, but from hands-on practical learning that is delivered through high quality polished content. Such practical training effectively complements all the theoretical content you learn in universities. Q-CTRL's Black Opal product is the world's leading interactive learning tool that gives you just enough fundamentals (think 101) required to program a quantum computer. With over 28,000 users worldwide and success across the UK (Tamil Nadu adopted it last year as part of the naanmudhalvan program). And now, we have expanded our presence in India through a network of local resellers. We encourage Indian companies, governments and universities to work with us either directly or through our contacts. https://lnkd.in/gjJWHYB3 We also look forward to aligning with AICTE standards and to make our content even more relevant to India. Waseem Shiraz, Shobhit Gupta, Nagendra Nagaraja, Reena Dayal, Anil Prabhakar, Kamakoti Veezhinathan, Ajai Chowdhry, Harshan Budke, Sunil Gupta, Arindam Ghosh, Ravi Puvvala Dr M Jayaprakasan, ISDS., National Skill Development Corporation, MindTech, Kquanta Research, RV College Of Engineering, Sanjay Chittore, Uttkrist Innovations Private Limited, Subu Gupta, AICTE QRDLab - Quantum Computing in India, IBM DRDO, Ministry of Defence, Govt. of India, Ashutosh Sharma, Urbasi Sinha, Bloq Quantum LTIMindtree, Tata Consultancy Services Abhay Karandikar, NPTEL, National Quantum Mission, Eltech, Andhra Pradesh Information Technology Academy ITE and C Department Government of AP, Jayesh Ranjan, CDAC Bangalore, Kaushalya The skill University

50

2 Comments

Buy side financial services companies (wealth / asset mgmt) have been striving to build/ scale AI solutions to deliver tailored autonomous financial advice, drive advisor productivity and improve operational efficiency. The following blogs complete few of my ideas around implementing / scaling AI. #3 Architecture Design - https://lnkd.in/gdXpG_zb #4 From Prototype to Production - https://lnkd.in/gKxNjdiD #5 Transformative Use Cases - https://lnkd.in/gaEsYiev (continuing from previous blogs for #1 - AI - Strategic Frontier, #2 - AI - Strategic Roadmap)

16

⚠️ Founders - consider the below reasons to not use Forward Deployed Engineers (FDEs) to acquire your *first* customers. This signals (1) lack of PMF; (2) miscalculation b/w development and sales timeframes; and (3) mischaracterization of Palantir Technologies timing when they deployed their FDE model. 💡 At the early-stage, it's crucial to demonstrate ability to repeatedly acquire customers with limited resources - capital, team, & time - that supports the narrative of 'more funding will amplify my traction.' Bespoke solutions for one company demonstrates zero repeatability. These custom solutions will not carry over to other customers within the same cohort. 💡 AI accelerates engineering but sales, system integration, and stakeholder buy-ins are perpetually slow. This imbalance results in concentrating all your engineering resources into one company, ignoring other opportunities, and still not going live on an accelerated timeline. 💡 Palantir didn't start using FDEs until year 3 when they got their first *gov't* customer, not their first customer. Gotham was already built and had strong PMF. Gotham allowed Palantir to take GTM risk by using FDEs b/c if it didn't work, Gotham was humming in the background. TLDR: At the early-stage, repeatable customer acquisition is the holy grail. Don't let AI deceive you that it can accelerate human behavior like it has with development cycles. It C.A.N.N.O.T.

33

1 Comment

Dashboards don’t create accountability. They create meetings. Charts get shared. Everyone nods. Action items get “noted”. Same problems. Next quarter. Accountability doesn’t come from visibility. It comes from systems — where ownership is clear and decisions can’t be ignored. I wrote a short note on why dashboards feel productive but rarely change behaviour — and what actually does. If your org has great dashboards but the same issues every review… this will sound familiar. Dashboards Don’t Create Accountability — Systems Do: https://lnkd.in/gCGpm-VU

7

1 Comment

YellowDog.ai just set a 10x benchmark uplift in scale computing, delivering 40,000 tasks per second (TPS) and managing 100,000 compute nodes in the cloud. What's even more interesting, that's 2x IBM Symphony and opens an intriguing pathway for these until-know closed/captive systems.

11