Anshu Aggarwal

Ever since we launched SHAKTI, every quarter we write a letter to our LPs = Each letter includes an update on the portfolios, the GP's (our) view on the zeitgeist, an interview with a portfolio CEO and an interview with a TITAN = So far, we have 16 such letters = it's been on our "to do" list to share these publicly (i am lazy) = then Google comes along to save lazy people like me (thank you NotebookLLM) = here's a 10 min podcast summarizing our journey since Dec/2020 with the 16 LP letters as source = if you wanted to know what we are about, this is a pretty good place to start = https://lnkd.in/db7K2TrT cc SHAKTI

66

5 Comments

Instacart is a market leader when it comes to leveraging AI tools to boost productivity at work. I had an amazing conversation with JJ Zhuang, Instacart’s Chief Architect, who has shepherded over 50 AI-driven initiatives to market across the org. We explored what catapulted Instacart to the forefront of AI integration, supporting radically diverse use cases across various functional areas. One striking insight (echoed by Ethan Mollick and others elsewhere) is that you often don't need specialized tools for specialized work. Frontier models like GPT-4 can seamlessly adapt to a wide range of functions and use cases. They're not just for engineers; they're incredibly versatile tools used by marketing, comms, and legal teams. From generating code to proofreading and ideation, GPT-4 empowers every department to leverage AI for enhanced productivity and innovation. By integrating LLM’s like GPT-4 into their workflows, teams can streamline processes, automate repetitive tasks, and unlock bandwidth for new creative possibilities. How can you harness the power of AI in your own work to boost productivity? Share your thoughts below 👇 Want to hear more insights from JJ? Check out the full podcast episode by clicking the link in the comments.

8

2 Comments

Batch processing for LLM inference activity: "One sign of that is the rise of batch processing. Today, popular consumer AI products like ChatGPT and Perplexity provide users with near instantaneous responses, otherwise known as real-time inference. However, businesses don't always need immediate responses and are often willing to wait hours, days, or even weeks for responses, as long as they don't have to shell out as much money. Founders of cloud and inference providers have told me there's growing pressure from business customers for this kind of flexibility, otherwise known as batch processing." -- Nathaniel Whittemore on "The AI Daily Brief" (with much deserved shout-out to his inspo from Stephanie Palazzolo at The Information) I fall into this bucket as we have a lot of backfilling to do as we're building out the underlying data in our service, and some of that could just as well be done in the wee hours of the night. The new stuff flowing in I want processed ASAFP, but these backfills—I'd mos def take more cost-efficient delayed batch processing. There's also use cases like needing (near) real-time inference for the first inference pass for timeliness / contemporaneousness value, but then a second pass slow-time inference for wider classification exercises over an ever-growing data corpus. I could see us mixing Sonnet & Haiku models in both cases. Just a heads up to Anthropic Mike Krieger as you contemplate this, along with Amazon Web Services (AWS) Jeff Barr Mark Roy Swami Sivasubramanian on the Bedrock side of things...and I don't mean the "batch" mode that doesn't seem to have any real value other than helping AWS manage its own rate limit challenges. (DeepCast is using Bedrock.)

11

2 Comments

During yesterday's LibreQoS APNIC webinar, I posed the question: won't more bandwidth solve these problems? As Herbert Wolverson said, yes, more bandwidth is good, but, still these problems remain if queueing is done incorrectly. Here's my take on why bandwidth alone is not as good as bandwidth + good queueing policies: Generally it was believed that 'data is important, so don't throw it away; hang on to it and send it later'. In practice, this has proven to be suboptimal as two issues may emerge: 1) latency increases for some flows due to heavy demand from other flows using the same bottleneck link, 2) even a single flow can have excessive latency due to aspects of typical TCP behavior (referred to bufferbloat) when the buffers grew large enough that the data being buffered was retransmitted anyway. It turns out that not all data is equally important. Active Queue Management is the art of deciding priorities, both in deciding what data to throw away, but also in allowing some later arriving data to be transmitted ahead of data in another connection that arrived before it. These problems have been studied, and good solutions have been found by using certain queueing policies in routers and switches, referred to as “fq_codel’ and “cake”. These track the different flows not by classifying the data, but by watching the behavior. Flows that send relatively little data (DNS lookups), or at a measured pace (video chat) have priority over flows that send a lot of data as quickly as possible (App and System updates, Video and ISO downloads).

10

5 Comments

DeepCast shipped a number of really cool new features recently in advance of a broad outreach push to podcast listeners. One new feature is our new Daily DeepDigest tool that sends you a concise daily email including any new episodes from the trailing 24 hours from any podcasts you Subscribe to (on DeepCast). Think daily podcast cliff notes or podcast news ticker tape, with a hint of Morning Brew various brews or PitchBook News—it's increasing information throughput when you're pressed for time. And this feature is rapidly evolving to distill other key extractions from shows you're already interested in, but also exposing you to other shows you might be unfamiliar with that also have nuggets to share—and thus increasing your aperture yet still conscious of your time. Want to go deeper on any given episode? Click the episode detail and you're off to an episode on DeepCast, where you can subsequently read a longer form distillation of that episode or go straight to the source and listen to the episode. (And podcaster friends, these pages are definitely increasing indexing and thus Google entry points to your shows!) I'm loving this feature as a user, personally. Can't wait to see what others think! #BuildingInPublic #Podcasts #AllTheWorldsInformation #HowYouWantIt

24

1 Comment

"Editors note: Welcome to issue #2 of Transcriptions. Today, we’d like to present a conversation between Qasar Younis of Applied Intuition, and Bilal Zuberi of Lux Capital. We discovered Qasar recently, and find his focus and clarity of thinking to be admirable. What follows is a lightly edited transcript of their talk, first held at the Paklaunch forum." Link to transcipt: https://lnkd.in/gH4uBmPE

12

2 Comments

Today, I was reminded of two remarkably prescient 1997 ACM, Association for Computing Machinery presentations: the first by Nathan Myhrvold and the second by Bran Ferran. If you are not a member of the ACM - I highly recommend joining. Nathan Myhrvold The future of software, the software industry, and Windows '47 https://lnkd.in/egMwYty Bran Ferran How IT will transform the experience of telling and listening to stories https://lnkd.in/gqF7C7TK

3

The magic of AI will only be unlocked for the agtech ecosystem once we have good ag data to feed the beast. We need digitized, harmonized, standardized, interoperable, machine-readable data for the magical AI tools to do meaningful work in agriculture. Let's get to work!

17

5 Comments

Here are the Top 10 SaaS funding rounds of the past week (Seed to A) Starting at number 1... 🥇 Emergence, $97.2mm Seed 🤯 Emergence is developing an AI "agent-based" system that automates tasks typically performed by knowledge workers. Founded by Satya Nitta. 💸 Investors: Learn 🔗 https://bit.ly/3XSitTA 🟠🟠🟠 🥈 Formic, $27.4mm Series A Formic enables manufacturers to scale faster by making it easy to deploy robotic automations, like packing and palletizing. Founded by Misa Ilkhechi and Saman Farid. 💸 Investors: Blackhorn Ventures, FJ Labs, and more 🔗 https://bit.ly/3XGzEYr 🟠🟠🟠 🥉 normai, $27mm Series A Norm Ai is automating compliance processes, making them more efficient, cost-effective, and accurate via (you guessed it) AI. Founded by John Nay. 💸 Investors: Coatue, Bain Capital Ventures, and more 🔗 https://bit.ly/4ckpNft 🟠🟠🟠 🎖️ 4. ORA, $20mm Series A Ora is a blockchain project incorporating AI into dapps. Founded by Kartin Wong. 💸 Investors: Polychain Labs, HF0, and more 🔗 https://bit.ly/3zpW7Ph 🟠🟠🟠 🎖️ 5. Payabli, $20mm Series A Payabli builds the infrastructure that allows companies, specifically software companies, to embed and facilitate payments through APIs. Founded by Joseph Elias Phillips and William Corbera. 💸 Investors: QED Investors, Bling Capital, and more 🔗 https://bit.ly/4bwkGHE 🟠🟠🟠 🎖️ 6. Sift, $17.5mm Series A Sift develops tools to accelerate the development and ensure the reliability of complex machines (think Spacecrafts). Founded by Austin Spiegel and Karthik Gollapudi. 💸 Investors: GV (Google Ventures) 🔗 https://bit.ly/3L7zlyk 🟠🟠🟠 🎖️ 7. Ario, $16mm Seed Ario is an AI-powered assistant for parents that helps with daily tasks, from meal planning to scheduling playdates. Founded by Sumit Agarwal. 💸 Investors: Wing Venture Capital, Floodgate, and more 🔗 https://bit.ly/4eEjeG9 🟠🟠🟠 🎖️ 8. MagicSchool AI, $15mm Series A MagicSchool provides lesson-planning solutions for teachers to fight burnout. Founded by Adeel Khan. 💸 Investors: Bain Capital Ventures, Adobe Ventures and more 🔗 https://bit.ly/3RQt5Pn 🟠🟠🟠 🎖️ 9. AuthZed, $12mm Series A Authzed makes it easier for developers to build permissions in their applications. Founded by Jacob Moshenko and Jimmy Zelinskie. 💸 Investors: General Catalyst, Work-Bench, and more 🔗 https://bit.ly/4bqkh9B 🟠🟠🟠 🎖️ 10. SLUSHY.com, $10.2mm Seed SLUSHY is a creator monetization platform. Founded by David Gross and Fred Sprivock. 💸 Investors: Mantis Venture Capital, Jon Oringer, and more 🔗 https://bit.ly/4eN9Boi ⚫⚫⚫ I'm Gyan, Head of Community at Venture House. I create tools, guides, and other free resources for Junior VCs and founders. If you found this helpful, check out the Venture House website for more 👇 https://venturehouse.xyz/ #startups #venturecapital #funding #saas

40

1 Comment

Let's finally get rid off the "VC alphabet song" and redo the system to become more founder friendly. "The benchmarks set by funding stages are often arbitrary and not necessarily aligned with the actual development needs of individual startups." #vc #venturecapital #founder #fundraising #startups #fundingstages #funding #fundingrounds #pitchbook

5

I’m excited to share this one! Redactive has raised a $11.5M seed round -- their mission is to become the trusted GenAI engineer for enterprise teams. Blackbird co-led this round alongside Felicis with participation from Atlassian Ventures and Zapier. Andrew, Alexander and Lucas are working to solve an urgent problem hidden in plain sight – empowering GenAI applications to meet their promise of "magic" in an enterprise setting. GenAI applications are only as useful as the data that they have access to, and for most enterprises, the most valuable data is sensitive and permissioned. We all want a chatbot (“ChatGPT for…”) at work that can search the data that we share in private Slack conversations or store in private folders on Sharepoint but first, we need to trust that our data will be kept safe – from prying eyes and accidental disclosure to others in the organisation who should not see it. Redactive keeps your sensitive data safe, giving developers the trust to connect it to GenAI applications and build products that you will love. Here's to the next decade. https://lnkd.in/gPuEAFvW

203

6 Comments

Insightful breakdown from my great Partner Omri Amirav-Drory on the simple factors that really separate the top 1% of investors from the rest, and how more operators should become investors! https://lnkd.in/gEiZdT2h

15

3 Comments

#hiring If anyone knows an awesome engineer that would be interested in joining an emerging #AI company doing important work in #healthcare, please drop me a line! Andy AI is backed by General Catalyst, YCombinator, Greater Ventures, and others, and they’re looking to hire a full stack engineer (US based) who's excited about using AI to improve healthcare and reduce nurse burnout. Andy is AI software that automates clinical documentation for nurses who do home visits. A nurse makes an audio recording of the visit and Andy automatically generates a form that usually takes nurses 90 minutes to complete. Nurses love the product because it potentially saves them 15 hours of charting a week! The founders previously built healthcare AI at Google Health and Apple, and are now building their vision to transform home health. More about Andy AI: https://with-andy.com/ Nurses love it: https://lnkd.in/g9XqZrUq #GenAI #digitalhealth #hiring #engineers #healthtech

25

2 Comments

LLMs are trained on internet-scale data. But to make AI work for work, models need to be grounded in your enterprise data -- a much harder problem than meets the eye! How do you connect to all your data sources and ingest them frequently enough while managing compute/storage costs? How do you make sure that you only show users results from documents they have access to? How do you rank results without internet-scale click and link data? How do you process the content of documents that are often hidden in PDFs and presentations? These are tricky problems, and it's why enterprise search didn't work for nearly a decade, until Glean came along. Turns out it's the exact same set of challenges needed to give LLMs the right context to work their magic on your enterprise content. Glean spent 5 years building the ingestion, search, and ranking infra needed to solve this problem well. This has given them the right to run the "Work AI" race and delight users at work. CEO Arvind Jain shares more on the latest episode of Training Data. Listen here, or wherever you get your podcasts: YouTube: https://seq.vc/ytd Apple: https://seq.vc/pit Spotify: https://seq.vc/jnz 

194

9 Comments

RethinkX's recent publication on The Rise of Human Robots and The Future of Work is quite interesting. Tony Seba and his team including Adam Dorr and Bradd Libby, do an excellent job dissecting the emerging humanoid robot technology market. Here's a brief summary, with a more in-depth summary from Cern Basher, CFA, an expert on the future financial impacts of humanoid robots, linked in the comments below: The rise of humanoid robots is poised to disrupt human labor across industries, much like how automobiles disrupted horses. This convergence of technologies creating a "labor engine" will enable robots to eventually perform virtually any task better and cheaper than humans. Initially, robots will complement human workers by taking on simpler tasks. However, their capabilities will rapidly improve until they can do almost anything humans can by the 2040s. This disruption represents both a massive opportunity and challenge for society. As the marginal cost of robotic labor approaches zero, we can expect sweeping deflationary pressures and quality improvements across the economy. Labor supply will explode, uncoupled from population growth, giving nations with robust robotics industries a strategic advantage. While creating prosperity, this transition will destabilize conventional economics and the social contract around work. Thoughtful policymaking will be critical to avoid catastrophic unemployment or stifling innovation through misguided protectionism. Ultimately, humanoid robotics could lead to superabundance and liberation from labor. But achieving this will require rethinking economics, embracing the disruption responsibly, and prioritizing policies that protect people over obsolete jobs and industries. https://lnkd.in/gjNDVQe4 #humanoids #robotics #futurefrontier

3

1 Comment

Stanford University's latest Foundation Model Transparency Index (FMTI) includes Writer LLMs, and the results are impressive! This was the first time Stanford Institute for Human-Centered Artificial Intelligence (HAI) evaluated Writer’s Palmyra-X model in FMTI, reflecting our growing impact to building customer trust through transparency, especially in enterprise applications where security, accuracy, and data protection are paramount. Palmyra-X achieved some of the highest marks for enterprise-grade models based on a set of 100 indicators. But in so many ways, our commitment goes far beyond FTMI expectations. At Writer, we understand that enterprises need top-notch performance paired with solid assurances about data management and protection. Our strategy isn't just to meet industry standards — it’s to surpass them. 🔍 Here are just some of the additional ways we're setting higher standards with our customers: Zero data retention: We never store data longer than is needed to provide our service, and give customers the right to delete data at any time. No training on customer data:
We don’t use the data you share with us to create, modify, or train our models. Zero model distilling & quantization: We do not distill or quantize our models, so customers can achieve peak reliability of accuracy, quality, and performance at all times. Toxic filter transparency: We share the mechanisms of our toxic content filters, ensuring our customers understand the safeguards we have in place to prevent harmful content generation. Bias mitigation: We’re transparent with our customers about our ongoing efforts and methodologies to identify and mitigate bias in our models. We’re providing our customers with technology that's not only powerful but also secure, and empowering them with the knowledge to make informed decisions. Our participation in this research is just the first step. We’re on a journey to raise the bar on transparency and redefine industry standards. Stay tuned for more to come! Link: https://lnkd.in/g8sMU2Qx

218

7 Comments

🚰 VODA.ai is building software to make one of our scarcest, most important resources safer and more cost effective to manage🚰 It's W13 / Q2 / 2024 & this week in our MGMT Boston startup series we’ve got VODA.ai, an AI startup helping analyze our nation’s water infrastructure at a local level to deliver better service for utilities and local populations across the country. Co-Founder & CEO George Demosthenous was working in product management at a utility technology vendor that builds metering & network infrastructure for utilities. George consistently heard that while the technology Mueller was selling was effective, the external pipe detection equipment and sensors they were purchasing didn’t necessarily stop leaks. There are 2.2M miles of drinking water pipes across the U.S. In Jackson, MS more than 65% of their processed water does not make it to end customers. In Detroit’s Highland Park neighborhood an estimated 70% of water is lost from 120 year old pipes. Imagine if Apple shipped 100M iPhones stateside and 50M were littered across the Pacific Ocean floor on their way over from Asia. Crazy, right? VODA.ai is building a software company that uses existing data to assess the condition of pipes that cannot be seen and determine which ones will have the highest likelihood of an event, like a leak or a break. VODA.ai created an AI engine dubbed “daVinciTM” here in Boston that uses pattern recognition in the way that LLMs probabilistically identify the next word in a sentence. daVinciTM has been tested head-to-head with other alternative models and has always outperformed! No pressure.. Since COVID they have grown both revenue & customer count by 200%+ y/y, continuing to leverage AI with both their product & team. AI’s emergence has helped VODA.ai’s sales cycle fall from a full year to three months as utilities broadly want to leverage best in class AI software. This Seed-stage startup has already analyzed over 1M miles of pipe (almost 50% of America!) and over 2M miles of service lines with customers from Boston to San Jose, CA. They are planning to more than 2x revenue in 2024, becoming the go to decision support platform for water utilities across the country in the next five years. VODA.ai is backed by local Boston investors Innospark Ventures and the Massachusetts Clean Energy Center. MassCEC has named VODA.ai a top carbon reduction emission leader, the team has twice been voted (by water utilities!) the Best Smart Water Solution at the Smart Water Summit, and VODA.ai plans to continue to grow their team in Boston in the quarters ahead. Operators to Know: -Kyra Dubinsky, Chief of Staff -Cory Sides, SVP Sales -Dan Hack, EVP of Operations and International Sales -Lowell Rust, MBA PMP, VP of Customer Success If you're interested in learning more, the full post is linked in the comments! And sign up for the newsletter to see who we bring you next week, will you?? #bostontech #startups #AI

43

3 Comments

After Apple’s research saying LLMs can’t think, it’s funny that now all I’m seeing is research on teaching them to think – something most humans still struggle with. Papers: Looking Inward: Language Models Can Learn About Themselves by Introspection https://lnkd.in/gf6rgnUP On The Planning Abilities of OpenAI's o1 Models: Feasibility, Optimality, and Generalisability https://lnkd.in/gNsPtDbi Thinking LLMs: General Instruction Following with Thought Generation https://lnkd.in/guQtRnsD #AIResearch #MachineLearning #ArticificalIntelligence

24