Danfeng Xu

Huge congratulations to the LanceDB team, Chang She and Lei Xu, on securing an $8M funding round from CRV! Their mission to build The Database for Multi-modal AI is an exciting one. It's amazing to see how far they've come since Chang's days as a star VP Eng at Tubi. Check out the TechCrunch article for more details on their journey. #AI #technews #fundinground

103

1 Comment

One interesting thing about Deepseek, according to its company CEO Liang Wenfeng, is that unlike many Chinese AI companies, they didn’t just copy Llama & tweak a few things, they innovated on a new architecture which allowed them to achieve these prices. Interview linked below.

7

1 Comment

Earlier this year, Jensen Huang - Founder / CEO of Nvidia - opined the following when asked where he thought the next amazing revolution was going to come from: - (Text transcribed verbatim) ***************** "This is going to be flat out one of the biggest ones ever, and I told you I was going to tell you, and I'm going to tell you. There's no question that digital biology is going to be it. For the very first time in our history, in human history, biology has the opportunity to be engineering, not science. When something becomes engineering, not science, it becomes less sporadic and exponentially improving. It can compound on the benefits of the previous years, and every company's contributions, every research's contributions, compound on each other. For the very first time, we know how to represent biology, understand the language of biology. We can represent the language of chemistry. It's never happened before. And now we're going to have incredible tools that bring the world of biology, which is very chaotic and constantly changing, and very diverse and complex, and bring it into the world of computer science. And that is going to be profound. And so if you happen to love this intersection, I think it's going to be rich with opportunities. It's going to be a giant industry." ***************** Now if you take this thesis and apply it to the world of #ClinicalDevelopment / #ClinDev in #LifeSciences, the opportunities to leverage #AI cuts across: * optimizing trial design * patient recruitment and retention * data analysis and prediction And the impact on the speed, cost, and efficacy of new therapeutics will be profound. And so I am really looking forward to joining some of the best minds in the intersection of AI and Life Sciences including Adam Petrich, MD | Arnaub Chatterjee | Prof Ava Amini from MSFT | Charbel Haber from Moderna | Fred Martin | Ittai Harel | Thaminda Ramanayake | Jasmine LaCoursiere | Gaurav Singal, MD former Chief Data Officer from Foundation Medicine | Orr Inbar | Sean Hilscher | Thorsten Wirkes from Bertlessman at the world's first AI Summit focused on clinical development in Boston on May 30th. Together we hope to demystify the role of AI in Clinical Development, discuss the risks and regulatory issues surrounding this field, and real-world applications and success stories, and the venture capital investment happening in this area. Apply to join us on May 30th - https://qhsummit24.com/

75

9 Comments

What I learned from Jerry Yang & Larry Page TL;DR: Technology companies launching disruptive products should religiously focus on end-users above everything else, including partners. Solve end-user problems and everything else follows– adoption, revenue, and partners. Yahoo! & Jerry Yang (2006) In 2006, phone calls were expensive and based on time & distance pricing. Skype changed all that with VoIP, which bypassed telco networks and made super cheap calls using the Internet. They already amassed 34 million users and were growing at insane rates. At the same time, Yahoo Messenger had over +110 million users worldwide. We could leverage this userbase and create a VOIP product to challenge Skype. Yahoo had all the advantages on paper: the brand, the users, and the technology. So, with the leadership of Brad Garlinghouse, we built a competing product with a very disruptive pricing plan – 1 cent/min; Skype's pricing ranged from 2-3 cents/min. However, a week before the launch, Jerry Yang paused the project at a partner’s request. ATT-SBC was concerned our pricing would cannibalize its own voice business. Yang did not want to upset that relationship. Ultimately, Yahoo missed its opportunity, and Skype was off to dominate the VoIP market. Google & Larry Page (2008) Before the iPhone, companies had to negotiate with mobile carriers to get their apps on phones. Google was struggling to secure deals with Verizon for Google Maps and a new phone concept, Android. Around the same time, the FCC planned to auction 3G spectrum. Google pushed for open application and open device requirements on the carriers who won the auction. This infuriated the carriers. They argued that such conditions would decrease the spectrum's value below the $4.8 billion auction floor set by the US government. The FCC was leaning towards the carriers. The agency did not want to risk the auction not meeting the government's financial goals. Google's only option was to participate in the auction to ensure the minimum bid. Inside Google, the decision to bid was intensely debated. Android founder Andy Rubin argued the carriers would hate us and never partner with Android; Chris Sacca argued that bidding was the only way to guarantee Google's future in mobile. Page sided with Sacca. Page’s rationale: “. . . the carriers already hate us. If we don’t bid, they would hate us and not respect us.”  We bid, the openness provisions were triggered, and we secured a path for Google apps on 3G. Sixteen months later, Andy got his deal Verizon –  Verizon Droid! Page was willing to sacrifice mobile carrier deals to ensure Google’s products were universally accessible to its users. He played spectrum poker with $4.8 billion to achieve that result. (cont. below)

29

11 Comments

From YC W24

30

4 Comments

Thanks to Youngjin Kim and Moweon Lee, two experienced Naver engineers with years of expertise in big data analytics, for sharing how they use StarRocks in complex real-time data analytics scenarios. Youngjin highlighted the reasons why Naver chose StarRocks: • 🚀Strong JOIN Performance • 🚀Smart Cache Strategy • 🚀Cloud Native Storage • 🚀Analytics with Flexibility • 🚀Materialized View To learn more, check out this impressive video: https://lnkd.in/gHu-WZND

7

I'm excited to host the AI4D: ByteDance Open Source Meetup on Thursday, October 3rd, from 5:30 PM to 8:30 PM at 1199 Coleman Ave, San Jose, CA 95110. Join us for a gathering of AI professionals to discuss challenges and solutions in AI implementation. Topic: AI/ML Inference Our keynote speaker is Dr. Alex Wei Chen, Founder and CEO of NEXA AI and creator of Octopus v2, an AI model that's 4x faster than GPT-4 and 70x more energy-efficient. He'll introduce Nexa SDK, an on-device AI toolkit supporting various models and GPU acceleration. Agenda: - 5:30 PM - 6:00 PM: Networking & Registration - 6:00 PM - 6:30 PM: Welcome and AMA with Dr. Alex Wei Chen & Tina Tsou - 6:30 PM - 7:30 PM: Nexa AI Presentation - 7:30 PM - 8:30 PM: Networking Dinner and refreshments will be provided. Space is limited, so RSVP is required. Please remember to bring your photo ID. If you're interested in presenting your company's product, contact Gary at [email protected]. Follow ByteDance Open Source on Twitter: [@ByteDanceOSS](https://lnkd.in/g5PDdYyY) Looking forward to insightful discussions and networking with leading minds in the AI industry! #AI #MachineLearning #OpenSource #Networking #ByteDance ByteDanceOSS

33

#AIComputeEngine As we move from the cpu centric era to accelerator era. Key Trends 1. Multimodal Data 2. AI powering AI 3. Compound AI 4. Scale 5. Reliability This is What we building AI Infrastructure Engine from last 3 years - NexaStack https://www.nexastack.ai #raysummit

18

1 Comment

Can we trust machines to judge machines when it comes to LLMs? There's an ongoing battle between manual and automated evaluation of LLMs. Manual evaluation is predominantly used by many companies, especially large ones with sufficient budgets. This approach involves internal QA teams or external consultants who manually create test cases and evaluate the outputs in spreadsheets. This process is time-consuming, but can provide insights into: - model's performance - specific failure modes such as hallucinations - tone discrepancies - biases - copyright violations - privacy issues Smaller companies without such budgets often resort to using expensive engineering time to manually review outputs, which can be inefficient. The primary advantage of manual evaluation is the accuracy and reliability of human judgment. But it's not scalable and introduces human biases, which can affect the consistency of the evaluations. Also manual processes cannot keep up with the rapid development cycles and the volume of outputs generated by modern LLMs. Automated evaluation aims to reduce the reliance on human labor by employing metrics and models to assess LLM performance. Techniques used in automated evaluation include: - Metrics-Based Evaluation: It utilizes established metrics such as BLEU score, ROUGE score, and perplexity. These metrics help assess aspects like fluency, accuracy, and relevance. But these metrics often do not correlate well with human judgments and can be inadequate for comprehensive evaluation. - Heuristic-Based Functions: These are specific functions to test particular capabilities or behaviors of the LLM e.g. toxicity detection or PII (Personally Identifiable Information) detection. These approaches can be tailored to specific use cases but may lack generalizability. - Model-Based Evaluation: It uses another trusted language model to evaluate the outputs of the target LLM. This can automate the evaluation process but requires careful calibration to ensure reliability. Automating this process can be a huge unlock. But developers attempting to create their own automated evaluation systems report very low confidence in the results. Lot of opportunity on this front.

3

1 Comment

The journey from Denny’s Diner to leading a 3 Trillion Dollar company: > born in taiwan > worked as a dishwasher at Denny's > went to Stanford > got a job > quit the job > went to california > started nvidia > pitched sequoia w/o a business plan > ventured into 3D chips > launched failed with NV1 chips > sold 250,000 to Diamond multimedia > 249,000 returned back to nvidia > NV1 ended in disaster > developed NV2 with Sega > NV2 was also doomed > had 1 shot to create the perfect chip > bought "emulator" from a dying company > made RIVA 128NV3 > nvidia engineers made the best inverse texture mapping engine > sold 1 million units > developed first programmable chip GeForce 256 > that was the world's first GPU > massive success > went public in 1999 > partnered with Microsoft to build XBox > didn't want to get trapped in the gaming market only > 2006, decided to put CUDA in GPUs > molecular dynamics, seismic processing, CT reconstruction, image processing were now open to explore > andrew ng came in with his ug student ian goodfellow to build a GPU server in his dorm room > 100x seedup in training neural networks on GPUs > Yann LeCunn at NYU as doing the same, all if them reached out to nvidia > Nvidia decided to put it's bet into AI which was in 2012 a 0 billion dollar market > nvidia reached to every single AI researcher in the planet > got positive feedback > continued building for AI > OpenAI launched ChatGPT > suddenly everyone was bullish on AI > 1999 stock price was <$0.5 > today, nvidia stands at $1220+ > a 298,534.15% increase > 31 years of nvidia > $3T in market cap in 2024

74

16 Comments

Check out the latest episode of our Tech Teams Today podcast with Yongsheng Wu, VP of Engineering at Granica - now available on all major podcast platforms. #AI #EngineeringLeadership #DataScience #MachineLearning #TechTeamsToday #Revelo

4

Every leader should watch this video. Thanks for sharing Nolan Church! In his book Loonshots, Safi Bahcall breaks down an innovation equation. One of the reasons I love this so much is that it recognises one of my favourite quotes from the late Charlie Munger about the power of incentives: “Show me the incentive and I will show you the outcome.” The innovation equation explains what creates innovation and it revolves basically around one idea. If employees gain the most value personally by delivering outputs which benefit the organisation, you’ll have high innovation. If they gain more value by doing anything else, you’ll have lower innovation as they will prioritise those things instead. Equation: Innovation rate = M = (E + S^2 + F) / G Equity Fraction (E) – how much equity does the team have in the success of an idea. This can be tangible (i.e. stock) or intangible (i.e peer recognition). More equity = more innovation. Fitness Ratio (F) – how well employees skills are paired with the work they do. Strong pairing means people are aligned to delivering value through their work (high innovation) rather than by lobbying for favour and playing politics (low innovation). Management Span (S) – The number of direct reports managers have (higher = more innovation, lower = less innovation). Salary Growth (G) – how your salary grows as your rank increases (higher = less innovation). If you look at this equation through the case study of Nvidia (undeniably one of the more innovative companies of our time) you see this playbook being run. - Low level of hierarchy, the Ceo has 60+ direct reports - Generous equity compensation with life changing outcomes for employees - Heavy focus on fitting skills to jobs Running these through the equation, you’d expect a highly innovative business! I love finding one element in a system which can drive an outsize gain in outputs. Especially if it’s easy to do. In this case, implementing a low hierarchy organisational structure is that one element. Taking out hierarchy and expanding the number of direct reports managers have, will dramatically increase the focus on outputs while reducing politics! This in turn generates significant value for the organisation because it focuses everyone on doing what is actually important. That this is literally the opposite of what most companies currently do, makes me like it even more. 🙂 Summary article on the innovation equation from the Harvard Business Review here: https://lnkd.in/eiNKH7fQ

20

1 Comment

How you can make your LLM applications better!

5

Building a Real-Time GenAI-Powered Food Delivery App with Redpanda Data Thanks Christina Lin https://lnkd.in/gHNfC-eA Harnessing the Power of Real-Time Data and Generative AI for Superior Delivery Experiences The food delivery industry is undergoing a rapid transformation, driven by the increasing demand for speed, convenience, and personalization. To meet these evolving customer expectations, businesses must leverage cutting-edge technologies like Generative AI (#GenAI) and real-time data processing platforms. Key Components of a Redpanda-Powered GenAI Food Delivery App: Real-Time Data Ingestion: Capture a continuous stream of data from various sources, including order placement, restaurant updates, delivery agent locations, and customer feedback. Data Enrichment and Transformation: Utilize Redpanda's in-situ data processing capabilities with WebAssembly (WASM) to clean, transform, and enrich data on the fly. GenAI Model Integration: Integrate advanced language models into the Redpanda pipeline to generate dynamic content, such as personalized recommendations, order summaries, and customer support interactions. Real-Time Predictions: Leverage AI to predict delivery times, optimize routes, and anticipate demand fluctuations. Continuous Learning: Continuously train and refine AI models using real-time data feedback to improve accuracy and performance. Benefits of a Redpanda-Powered GenAI Food Delivery App: Enhanced Customer Experience: Deliver personalized recommendations, real-time order tracking, and proactive customer support. Optimized Operations: Improve delivery efficiency, reduce costs, and increase driver satisfaction through intelligent routing and resource allocation. Competitive Advantage: Gain a first-mover advantage by offering innovative features and superior service. Data-Driven Decision Making: Make informed business decisions based on real-time insights and analytics. General Hashtags #Redpanda #GenAI #ArtificialIntelligence #MachineLearning #DataScience #RealTimeData #FoodDelivery #Tech #Innovation #Technology #DataDriven #DigitalTransformation More Specific Hashtags #RedpandaPoweredAI #RealTimeDelivery #AIforFoodDelivery #FoodTech #DeliveryOptimization #CustomerExperience #DataEngineering #WASM #TensorFlow #AIandBusiness Engaging Hashtags #FutureofFoodDelivery #TechTrends #Disruptor #InnovationThatMatters #TechForGood #SmartDelivery

9