< Previous30 edge_may 2024 interview I n 2022, Confluent, the global data streaming company, announced the Confluent for Startups programme, focussed on providing Confluent Cloud credits to early-stage startups that are working to get their streaming initiatives off the ground. By 2023, the company had worked with over 250 startups to help them put their data in motion. Building on this momentum, Confluent, along with global venture capitalists like Matt Miller, Sequoia Capital, Erik Vishria, Benchmark Capital, and Carlos Gonzalez-Candenas, Index Ventures, announced the launch of the ‘Data Streaming Challenge’, a global competition focussed on recognising early-stage startups. With a grand prize of $500,000 and two runner- up prizes of $250,000 each, the competition aims to inspire and help the growth of early-stage startups in the data technology landscape. Apart from providing startups with scalable, cost-effective, and a robust foundation to build their critical data infrastructure, Confluent, aims to work closely with these startups, to help them grow. In conversation with edge Tim Graczewski, Head of Confluent for Startups, talks about the programme, and its main aims. With over 25 years of work experience in technology in the Silicon Valley, Graczewski has held senior leadership roles in alliances, strategy, and business development Confluent Cloud provides startups with a robust and cost-effective foundation across companies like Intuit, AWS, and Oracle. Graczewski is also a two-time venture-backed entrepreneur (Nav and Cake) who has also helped raise over $100 million in venture capital on the way to building disruptive startups in fintech and consumer internet. Leveraging his unique mix of corporate business development leadership and hands-on entrepreneurial experience, Tim came to Confluent in January 2023 to lead and grow the recently launched Confluent for Startups programme. Give us the details of the Confluent for Startups programme, and how does Confluent help startups that participate in the programme? Confluent for Startups is designed as an on-ramp for promising early-stage startups (less than five years old) with great Kafka use cases. Eligible startups receive up to one year of Confluent Cloud for free ($20,000 maximum), along with access to on-going technical help and strategic design reviews. Confluent Cloud provides startups with a robust, scalable, and cost-effective foundation on which they can build their critical data infrastructure and solve their most data intensive challenges. Doing it the right way, from the beginning drives faster development, better insights, and a competitive edge in today’s data-driven business landscape. edge_may 2024 31 interview Confluent and our venture capital partners recognise the potential of new startups to revolutionise the world with data streaming applications, and we’re dedicated to helping them realise their dreams, whether that’s the next hit mobile app or helping the world’s largest enterprises monitor and reduce their carbon footprint. How did you and your team decide on creating the programme, what was the kind of work that went behind the scenes? Confluent understood that early-stage startups making the initial move to a new technology like data streaming requires a lot of deliberation around how to deploy their limited capital, people, and time. We designed Confluent for Startups to help early-stage companies ease the transition and leverage all the benefits of data streaming from the start. We wanted to demonstrate a deep commitment to the global startup ecosystem. We benchmarked with similar enterprise startups programmes to deliver the most valued benefits package possible. How does working with startups help Confluent and vice-versa? Startups benefit from Confluent for Startups in several ways. First, and most obviously, they receive free credits to build and test proof of concepts (POC) without cost. Second, they often receive expert feedback from Confluent’s solution engineers (SE), who have real-world experience designing and implementing sophisticated data streaming pipelines. This introduction to SEs can help startups “think bigger” about how they can scale their business their business. Finally, startups can increasingly learn from each other via our growing Confluent for Startups community, which is connected via a dedicated Slack channel. As for how Confluent benefits from working with startups, we get to see the new and innovative ways they are putting data streaming to work, such as this year’s Data Streaming Startup Challenge finalists, who have leveraged Confluent to power a computational chemistry platform (Atomic Tessellator), an operating system for group transportation (Busie), and the creation of digital twins platform for events and venues (TwinLabs. ai). Most large companies have their own startup accelerator/programmes, what differentiator does Confluent bring in? Many peer companies in the data infrastructure space (e.g. Databricks, Datadog, Snowflake), also offer free product credits via a startup programme. However, in addition to credits, Confluent for Startups offers a bi-weekly technical office hours hosted by a senior member of the Confluent technical marketing team, two one-hour sessions with an SE, and (for those that qualify and apply) a chance to receive a meaningful seed investment from Confluent (up to $500,000) – and pitch to some of the most prestigious venture firms in the world (Benchmark, Index, and Sequoia) – as part of our Data Streaming Startup Challenge. What are the kinds of startups that Confluent is looking at and why? Confluent for Startups is looking for all kinds of startups from all corners of the world that meet our basic eligibility criteria (less than five years old and new to Confluent). That said, we are particularly interested in several industries and use cases. For instance, we love to work with startups that started on open-source Kafka and are now looking to migrate to a managed service to benefit from the ease of use, improved performance, and lower total cost of ownership. Next, we believe that Confluent has a major role to play in the unfolding artificial intelligence revolution as new learning models move toward real time data and applications. Lastly, there are many established use cases that we know we can help young startups with immediately and significantly, such as fraud detection, IoT data integration, and real-time applications for inventory, business analytics, order management, field service, quotes, payments, and more. Tim Graczewski, Head of Confluent for Startups32 edge_may 2024 interview What are Confluent’s plans for the programme and its startup strategy in general? We have received almost 500 applications from 50 countries around the world since the programme launched in October 2022. We plan to continue to grow that footprint around the world in FY’24 by promoting the Confluent brand across the major cloud provider’s startup programme and the leading startup accelerators, as well as proactively educating the startup ecosystem on the power of data streaming and the value of Confluent for Startups. How does Confluent help startups refine their products’ MVP? How does Confluent direct startups towards building a scalable product? Confluent for Startups is designed to be an easy on-ramp to Confluent Cloud and data streaming for early-stage startups with real-time data needs. It is not a startup accelerator programme per se, such as Y-Combinator, Alchemist, or TechStars, that seeks to nurture young startups by providing guidance on such things as product-market fit and improving the narrative of their pitch deck. To the extent that Confluent for Startups provides coaching and guidance it occurs at the data architecture level. Each startup programme member gets two one-hour sessions with a Confluent systems engineer (SE). These sessions allow the Confluent SE to provide feedback on best practices for creating a scalable and robust data architecture that supports the startup’s needs and vision. The product is often a data architecture diagram that allows the technical founder to “see” how data is going to flow through their system. The $1 million Data Streaming Startup Challenge served a different purpose than our flagship Confluent for Startups programme. The Startup Challenge aimed at identifying and promoting early-stage startups that had already adopted Apache Kafka and Confluent Cloud and applied it in an innovative fashion to a novel use case. As the list of applicants was trimmed from 100 down to 12 and then down to three, we took a more hands-on approach to assisting the startups with their architecture, their pitch decks, and with introductions to possible partners. Confluent for Startups is designed to be an easy on-rap to Confluent CloudYour biggest source of industry insights SUBSCRIBE HERE EDITORIAL INQUIRIES e: editorial@industrymena.com SALES INQUIRIES e: sales@industrymena.com ENERGY AVIATION CONSTRUCTION TECHNOLOGYDEFENCEWALKING THE TIGHT ROPE OF GEN AI’S ENERGY APPETITE AND SUSTAINABILITY Can large language models find alternative energy sources to be sustainable? AI By Sindhu V Kashyap 34 edge_may 2024 featureedge_may 2024 35 feature36 edge_may 2024 feature Artificial intelligence and tech solutions, can be an energy drain O ver 85.4 terawatt-hours of electricity is used annually by small countries. This also could be the same amount of electricity consumed by generative AI (GenAI) engines, especially if its current adoption trends continue, according to research data conducted by the Scientific American, and Joule Magazine. If we add this energy consumption to the existing 1.5 per cent global electricity existing data centres use, it could put a crimp on those looking for a balance between AI and sustainability, especially in the MENA region. While AI is a powerful tool, there clearly are concerns about its energy consumption. However, like all things AI and technology, there is a double-edged sword element. Though an energy drain, AI and tech solutions can also help companies decrease waste and save resources with accurate mathematical calculations. “Though there is no precise number calculated on energy consumption of AI models, it is safe to say the consumption of data and information by these models is just going to increase. With these growing demands of genAI models, there is a need for systems that are not only efficient but also sustainable. By developing an energy strategy, organisations can also control costs and extract more value from data as it grows,” said Fred Lherault, CTO Emerging, Pure Storage. He explained that having smart distribution units and storage arrays, helps provide a precise measurement of energy consumption. “With this, companies can reduce server power usage and store data in powerful storage systems, as these are advanced enery management systems.” Lherault added that it also eliminates the need for several internal storage devices, reducing power consumption by up to 85 per cent on a terabyte basis. Pure Storage, for example, is already has this model in place. However, the focus on automation through AI solutions like demand forecasting is also a testament to its potential to reduce this footprint. These solutions can be used across industries. Alex Ponomarev, CEO of Syrve MENA believes by optimising operations and potentially streamlining the need for multiple AI models, they can contribute to a more sustainable food and beverages industry, for example. “Furthermore, our cloud-based platform leverages data centres that are increasingly adopting renewable energy sources. This ensures edge_may 2024 37 feature The training and inference phase are two energy consuming phases of AI a greener infrastructure to power our solutions,” commented Ponomarev. As explained by researchers, there are two big phases when it comes to AI – one is the training phase, where the model is being set up and taught how to behave on itself. This is followed by the inference phase, where the model is put into live operation and fed prompts so that it can produce original responses. The inference phase is particularly energy-intensive as it involves real-time processing and decision- making. “There is a complex interplay of trade-offs as we don’t have a full picture. Alternative energy sources are becoming drastically cheaper, and thanks to AI, the go-to-market for storage and distribution tech is contracting. However, AI energy demand may double by 2026,” added Amina Musaeva, founder of Cloudset. Some claim that superior autonomous intelligence would ultimately resolve and offset this energy waste, a classic techno-centric argument in a dilemma of progress versus climate change. Musaeva added that AI can consume more energy but can also make achieving results faster and be more efficient. Thus, this demand could be compensated by the energy savings on a consumer’s end. However, information on AI training is becoming more secretive, and these models are getting increasingly complicated. Along with this, there also is the problem of the hype around AI models. This induces many to throw more computing information and data as the systems and the hardware become more efficient. “A crucial starting point is to establish transparent reporting on the AI energy demands, create universal metrics for monitoring the offsetting and efficiency gains, and potentially introduce some energy capping and kWh trading strategies. This will ensure that only those with the highest need and greatest efficiency potential obtain the most quotas,” added Musaeva, emphasising the importance of transparency in AI energy consumption. However, all isn’t lost, new AI models are developed to count in on the energy factor. CLOVER by Cloudset, for example, is a specific AI model that adjusts its size based on the task at hand. It figures out what a user is trying to do and then selects only as big a model as that particular task truly needs. “The team reported that CLOVER can cut the greenhouse gas emissions of AI used at a data centre by more than 75 per cent. And with those savings, the accuracy of AI models’ results drops by only 2 to 4 per cent. Tech companies can route their data calculations to data places and data centres mostly powered by renewable sources. However, the supply is still too limited,” explained Musaeva. As AI engines start becoming ‘smarter’ and more efficient, the amount of data consumption and energy will only increase. This not only means more electricity for the AI models, but also for the data centres. Since the models are relatively new, the amount of energy they may consume isn’t known. Nevertheless, while the exact amount of energy these systems will consume is yet to be determined, many believe now would be a good time for regulators to start taking data on energy use discourses from AI developers to increase data sets and pools. edge smart buys 38 edge_may 2024 A deep dive into Appsflyer Creative Optimization Solution The AI-driven solution helps companies of all sizes unlock their advertising potential TECHNOLOGY By Sindhu V Kashyap edge_may 2024 39 edge smart buys I n January, AppsFlyer launched the ‘Creative Optimization Solution’, a new product that provides marketers with unparalleled insights into their creative assets and data-driven guidance on maximising impact. Coupled with artificial intelligence (AI), ‘Creative Optimization Solution’ identifies patterns, trends, and features that drive optimal audience engagement. This enables marketers to capture the most value from their ad spend, while enhancing the effectiveness of their creative content and campaigns. Adam Smart, Director of Product (Gaming) at AppsFlyer, explains that historically, advertisers and marketers have faced several challenges when it comes to effectively visualising and analysing the performance of their creative assets across various advertising networks. Traditionally, data aggregation has been at the campaign or ad set level, which makes pinpointing the impact of indivdual creatives cumbersome. This lack of granularity often leads to suboptimal decision-making and missed opportunities for optimisation. Moreover, manually tagging and tracking creatives across multiple networks introduces the risk of human error, further complicating the process. While data aggregation at the campaign or ad set level is standard, isolating the performance of individual creatives has been difficult. Manual tagging and tracking processes are prone to human error. Recognising these pain points, Smart’s team set out to simplify the process by providing a solution that offers comprehensive visualisation and analysis of creative performance. The idea for the ‘Creative Optimization Solution’ originated from recognising the increasing importance of creatives in the face of evolving privacy regulations and advertising landscapes. By collaborating closely with design partners and early adopters, the team refined the solution to meet the needs of marketers, user acquisition teams, and creative professionals. Through iterative testing and feedback, they validated the solution’s effectiveness in providing actionable insights into creative performance. Overall, the evolution of the solution reflects a commitment to addressing the evolving needs of advertisers and marketers in an increasingly complex and competitive advertising landscape. “The purpose of the ‘Creative Optimisation Solution’ is twofold: firstly, to address the inherent complexities associated with tracking and analysing creative performance across diverse advertising networks, and secondly, to empower marketers and advertisers with actionable insights to optimise their campaigns effectively. By leveraging AppsFlyer’s attribution capabilities, the solution aims to provide users with a holistic view of their advertising efforts, allowing them to attribute installs and engagements back to specific ads and creatives,” explained Smart. This level of visibility enables users to identify top- performing creatives, optimise budget allocation, and make data-driven decisions to maximise ROI. Creative Optimization Solution, an AI-driven solution to help marketersNext >