Arun Dobriyal

Worth reading: Generative AI is not going to build your engineering team for you. So, so good - peppered with gem quotes and observations- chuckle. https://lnkd.in/gZBKtU4X Thanks to Eric Brooke for sharing.

16

I am excited to share the latest blog, co-authored by myself, Junhwi Kim and Vyas Sekar, detailing how we managed to successfully localized and improved LLM by combining Fine-Tuning and RAG. https://lnkd.in/gKDkgGPq

70

3 Comments

Stripe just launched an SDK for AI agents enabling LLMs to ‘access financial services and earn and spend funds’. You can imagine this as part of an AI agentic workflow booking and paying for your trip, for example. Worth heeding the warning at this point: “This SDK is an early exploration of integrating Stripe into agentic workflows. As the behavior of agents is non-deterministic, we recommend exploring the SDK in test mode and running evaluations to assess the performance of your application. Additionally, we recommend using restricted API keys to scope access to the functionality your agent requires.” The more I learn, the more convinced I am this is the future we’re heading towards. “With the Stripe agent toolkit you can now easily integrate Stripe into the most popular agent frameworks. This enables you to automate common workflows that depend on Stripe and also helps unlock new use-cases by providing agents access to financial services on tools. In addition, usage-based billing can quickly integrate in these agent frameworks to bill your customers. Take a look at agent toolkit documentation to explore more of its capabilities and how Stripe is supporting agent businesses.” Lot to learn, and sort out, though. #genai #innovation #startups

24

6 Comments

A super clear, straight forward example of how to do cost effective RAG using LangChain for your document pipeline and Groq for compute resources. The code demonstrates how to speed up RAG and reduce costs using quantized embeddings (reducing 32 bit values to 8) Nice job Sahibpreet Singh! https://lnkd.in/gf5QMCKK

44

2 Comments

“AI-First” Design is the Future of Application Development It’s becoming clear that simply tacking on AI features to existing workflows is not enough to unlock their full potential. Just as mobile applications were transformed by the shift to a “mobile-first” design philosophy, today’s applications need to be reimagined from the ground up with an “AI-first” mentality. A decade ago, as mobile devices became ubiquitous, companies found that they couldn’t just retrofit desktop apps for smaller screens; they had to rethink the user experience for mobile. I see this same shift happening with AI. Companies that bake AI into the DNA of their products stand out, while those that merely bolt on chat and LLM features may find themselves falling behind. Take a classic example: forms. These are everywhere in application design, and most are terrible because they force the user to conform to rigid questions and formats. Consider e-commerce. If I want “a blue bicycle for a 12-year-old girl that’s really safe and has great ratings,” it takes me 10 seconds to type, but navigating through filters, clicking options, and reading reviews could take much longer. An AI-first approach would act as an agent on user needs and streamline that journey—transforming an exhausting task into a seamless experience. However, shifting to an AI-first design isn’t without its challenges. Existing workflows are often deeply ingrained in technical debt, making it difficult for companies to rethink processes at a foundational level. But those willing to disrupt the status quo will be the ones driving meaningful innovation. How to Get Started: Embracing an AI-first mindset doesn’t mean overhauling everything at once. Start small, but think big. Identify key areas where AI can have the most impact—whether it’s personalizing experiences, automating tasks, or improving decision-making. From there, build AI-driven solutions that elevate the entire user journey, rather than just tweaking existing features. A New Standard: AI-first design isn’t just an advantage—it’s the new standard. Companies that fully adopt AI into their product experience will create smarter, faster, and more intuitive products that resonate with users and set them apart in the market. This is more than a trend; it’s a shift in how we think about and build for the future.

40

3 Comments

One of the most interesting things to come out of of crypto - where the exchange of rewards is so fluid, and fraud is so rampant - is the creation of systems to define which users are real / not bots, and actually generating value. Airlines, hotels, and coffee shops have been in the loyalty business forever... but not with this kind of global reach or velocity. Everyone is already an influencer - but only some are big enough to get marketing contracts. Crypto is showing that anyone can get paid to help build or grow a product. https://lnkd.in/g_wci5tU

7

We are excited to back Morphing Machines. As Dr. Ranjani, CTO and co-founder says: “For the first time in India and globally, we have a processor that is truly runtime reconfigurable, catering to emerging workloads.” While Nvidia’s growth and high valuation are driven by robust demand for “infra-compute” in high-performance GPUs, spurred by Gen AI, there is a distinct need for innovative solutions tailored to a new wave of reactive applications. Enter Morphing Machines. It addresses this demand with its unique product designed for next-gen telecom, autonomous vehicles, collaborative robots, avionics, and online data analytics. This product is specifically engineered to handle heterogeneous workloads and diverse tasks with varying performance demands simultaneously, ensuring a high level of scalability. As we move from an institutionally-backed research team at IISC to productization, Golden Sparrow is collaborating with Speciale Invest and others to support Deepak Shapeti, who joined as CEO three years ago after earning his Masters at Stanford. He brings not only technical expertise but also the business acumen necessary to guide the product to commercial success. Venkat Raju Afreed Faizan Anand Rao Ritik Singh Michael Marmor

68

5 Comments

I like this idea of "technical taste". It gets into where software engineering is as much creative art as it is science. Four great takeaways (IMHO) for software devs in this era in particular: 1. Aspiring engineers should cultivate a sense of curiosity, experiment with different tools and technologies, and embrace a mindset of continuous learning. 2. AI has the potential to streamline processes and enhance productivity for engineers, but it may also lead to disruptions in traditional software development workflows. 3. Developing technical taste and judgment is essential for making informed decisions about which technologies and approaches to pursue. 4. Collaboration and open-mindedness are key to leveraging the full potential of AI and staying ahead of technological advancements. I also really liked these two quotes from Sam Schillace: 1. "The right time to do something is when you have that feeling in the pit of your stomach that's like, 'oh, this is a great idea and it's going to suck to build because nothing's ready yet.'" 2. "Technical taste is like, 'how well have you consolidated that set of experiences and heuristics into judgment that you can apply accurately when you see new things?'" 3. "It may be the case that very small teams can do very large projects, or like we were talking about before, it may be the case we're just going to get really ambitious about what we try to do with the same size of teams, which is kind of where I would put my money." Great job on continuing to put out great episodes, Brett Berson and team First Round Capital! #ai #engineering #softwaredevelopment

5

OpenAI Introduces Structured Outputs in API! #OpenAI's latest innovation: Structured Outputs in the API. This game-changing feature ensures model outputs precisely match developer-supplied JSON Schemas, revolutionizing how we build AI applications. Key Takeaways: 🎯 100% reliability in matching output schemas with gpt-4o-2024-08-06 🛠️ Two implementation options: Function calling and response_format parameter 🔒 Enhanced safety measures with new refusal string value 🐍 Native SDK support for Python (Pydantic) and Node.js (Zod) 📊 Perfect for generating structured data from unstructured inputs ⚡ Improved efficiency for developers, reducing workarounds and retries This update marks a significant leap forward in AI application development. By ensuring model outputs conform to specific schemas, OpenAI is addressing one of the core challenges developers face when integrating AI into their systems. I'm particularly impressed by the 100% reliability score in complex JSON schema following. This level of precision opens up new possibilities for creating robust, AI-powered applications across various industries. #openai #ai #artificiaiIntelligence #apidevlopment #structureddata #ml #machinelearning https://lnkd.in/g3tTyimS

This has absolutely been my experience building prod LLM-based apps. It requires a change in mindset from ‘this passes all the tests’ to ‘this passes enough of the tests often enough and is valuable enough that we’ll work around the times it fails’. It also creates some weird dynamics in terms of SDLC because as Brian correctly points out creating an exciting prototype is child’s play and whereas for normal apps the prototype might represent 50% or more of the work required to ship with LLMs the prototype can do maybe 80% of what you want but the effort remaining to get it to 99% effectiveness is 100x what it took to create the POC. We’re all having to learn new ways of thinking about dev if we want to work productively in this space.

10

2 Comments

Meta just released LLaMA 3 model which is an open source #LLM and you can now play with it on an OpenAI like interface that you can interact with at https://www.meta.ai Found this cool video so you can see how well #LLaMa 3 actually does in solving some problems you might need help with... https://lnkd.in/g6AtmG3h

8

What does the new MLE-bench tool from OpenAI offer AI developers in evaluating machine-learning engineering capabilities? "The new benchmarking tool from OpenAI does not specifically address concerns about the future of AI engineering systems, but it opens the door to developing preventative tools." OpenAI's MLE-bench tool assesses AI agents on 75 real-world Kaggle tasks, enabling evaluation of AI engineering performance.

5

Venture Capital Contact Compilation. Explore a detailed list of 1,250+ VC firms, 7,400+ professionals in the VC space (including partners and principals), and 6,400 email contacts. https://payhip.com/b/cs7UQ #startup #startups #vc #venturecapital #investmentbanking #privateequity #managementconsulting #b2b #founders #entrepreneurs

1

1 Comment

WIESNER ON HUMANS VS AI: MIT Media Lab co-founder and science advisor to JFK, Jerome B. Wiesner was a unique figure in the history of MIT at the crossroads of science, engineering, and the arts with a lifelong pursuit of world peace. There's a terrific talk he gave at the founding of the MIT "E15" building designed by I.M. Pei — often nicknamed "the bathroom building" — that's worth a watch, especially during these extraordinary times https://lnkd.in/g2zEJRrk and a piece of the talk he gave is below. Information on the talk I'm giving with Jerome B. Wiesner Professor Hiroshi Ishii is here https://lnkd.in/gJpKkeY8 on November 8th (Fri) from 11 am EST, and you can register online for free here https://lnkd.in/gADUyCs7 --- Jerome B. Wiesner at the founding moment of the MIT Media Lab's E15 building https://lnkd.in/g2zEJRrk > ... Now some people will be offended by what I'm going to say, but I think man-made systems use better components. Electrical signals travel a million times faster than signals on nerve fibers. And semiconductor devices also act billions of times faster than neurons, which are the biological information processing elements. Also, manmade elements appear to be more reliable. But nature has a very large edge, maybe a decisive edge. In design complexity involving a number of individual elements, it can use it in a system effectively. And these are advantages man may never match. Nature may always retain this edge, who knows? > But for the moment, it really doesn't matter. The manmade systems that we are involved with enhance human creative capacities, that's what they can do, but so far they cannot displace them. Cognitive and computer scientists are having an exciting time trying to understand and imitate human reasoning with computing systems. And though they can do a few things that would be regarded as thinking if done by a human, they differ violently among themselves on whether or not machines with human thinking and creative abilities are possible. Even the so-called artificial intelligence field, which prompted Walter Rosenblith to say he preferred natural stupidity, his greatest achievement in this field is capturing knowledge and skills of humans -- of human experts. > For now, it’s clear that thinking machines and true creativity don’t exist. We should focus on how tech can expand human creative capacity.

53

4 Comments

Re-posting one of my less popular / unorthodox views. Weekly 1:1s are hacks to get around the lack of proper team communication setups, and they weaken individual resilience. Adhoc 1:1s are fine for providing spot feedback, but they should be rare. Career growth 1:1s are great, but they should happen once in 3-6 months. Project check-ins / questions should be discussed in 5-10 minutes on slack. A design whiteboarding meeting between two engineers is fantastic, but it's not a 1:1. Weekly 30 minute 1:1s are unhealthy and an unnecessary time-sink, stop using them to communicate with your reports. https://lnkd.in/ehr_e2Xv

20

9 Comments

https://lnkd.in/gqUkq3BM is an excellent overview of major topics in LLMs. If you are getting started in this space, this video which is also generated by another LLM 😂 , is a great place to begin your journey in this exciting area of AI. It covers topics like pre-training, evolution of GPTs and other popular LLMs, fine tuning (SFT, PEFT like LORA, RLHF), inference optimizations (quantization, speculative decoding, distillation, flash attention), output sampling techniques, multimodal capabilities and finally some real world applications like math, coding, medicine etc. Loved their analogies even for core complex concepts like flash attention - "it's like finetuning the engine of a car, to make it more fuel efficient, without changing how the car drives"

24

India is not just catching up in the AI race; we're innovating and leading in ways that are uniquely ours. The Economic Times recently highlighted how our approach to Generative AI goes beyond just Large Language Models (LLMs), addressing the specific challenges and opportunities within our diverse market. Rahul Agarwalla, Managing Partner at SenseAI Ventures, shared his invaluable insights into this AI revolution. His leadership and vision are truly driving innovation in the field, and it's an honour to be part of this journey with him. For anyone interested in the developments in AI and how India is making its mark on the global stage, this is a must-read.

45

3 Comments

One of the main key choice points when doing GenAI app development is choosing your foundation model. Should you go “closed-source” frontier model, like GPT-4 or Claude 3.5, or try to stick with an open source model like Llama3? Or does your use case fit a more purpose-trained model like Gorrilla or Stability? This mostly comes down to testing, and understanding the economics of your app (assuming it’s a commercial effort) to decide which model to build upon. But waiting behind this decision is another key choice - choosing what level of *abstraction* on which to build. The simple approach is to use the native API provided by the model itself. This can be a good choice, but strongly limits your ability to support other models in your app. Prompt instruction syntax, tool calling syntax, and JSON or markdown generation can vary pretty wildly across different models. So introducing some abstraction to represent the LLM offers the possibility of supporting or swapping different models depending on your needs, without requiring heavy surgery to your app. Unsurprisingly there are lots of options vying to abstract your LLM interface, including LangChain, LLamaIndex, vllm and many more. Testing and picking amongst these is a project in and of itself. Generally you evaluate each solution across a few different axes: Does it support commercial or open source models, or both?  Does it look to support “advanced” features of different models, or rather attempt a “lowest denominator” approach to preserve compatibility? What other features (agent behavior, RAG, indexing…) is it trying to offer? How much will you be bound to those implementations if you choose this abstraction? I haven’t seen it yet, but this would be a great opportunity for a “bake-off” type evaluation to produce a matrix to help you understand the different options and trade-offs. Maybe I’ll go ask ChatGPT to write it up… #ai #genai #llm

23

1 Comment

AI can be a game-changer for recombining existing technologies and processing vast amounts of data, but it's not a magic bullet for radical innovation. 💡Companies that use AI to streamline and enhance existing processes often see benefits, but those aiming for breakthrough ideas might find it less useful. It’s a solid reminder that how we use AI, rather than the tool itself, determines its impact. 🚀 #AI #Innovation #TechTrends #BusinessStrategy https://lnkd.in/e5zTUDrC

7