Artikelen van Bamiyan
Activiteit
18K volgers
-
Bamiyan Gobets heeft dit gerepostBamiyan Gobets heeft dit gerepostGoldman Sachs predicts AI agent token use to grow 24x by 2030. Enterprise agents will soon generating more SQL queries than humans. OpenAI and Anthropic are partnering with Databricks and Snowflake because the foundation for enterprise AI agents is structured enterprise data. Speedata.io APUs can deliver 1-2 orders of magnitude better performance, price-performance and energy efficiency for SQL analytics compared to CPU and GPU.
-
Bamiyan Gobets heeft dit gerepostBamiyan Gobets heeft dit gerepostWe are living through three simultaneous transformations. The most consequential technology change since the birth of humanity. The most significant world order transition since the Cold War. And the largest industrial infrastructure built out in history. But these are not separate stories, they are all part of a single story. Now the key question at the center of it is whether open societies, the ones that require consent, debate, democratic participation, can build fast enough and fairly enough to stay sovereign. The biggest risk is not that we fail technically, it is not that we hit a wall, it is not even that China or some other rival outpaces us in raw capability. I think the biggest risk is that Western democracies look at this moment, see what is required of them -- the energy build out, the industrial mobilization, the policy reforms, the investment and the political courage to drive frontier Intelligence into the hand of every citizen - and they fail to rise to the occasion.
-
Bamiyan Gobets heeft dit gedeeldStrategy 1: Free up the servers you already own Strategy 2: Buy fewer servers for the new expansion Speedata's APU technology enables several strategies to repurpose existing server estates, or reduce required servers for new expansions. Either of these approaches offsets millions of dollars *today* in overpriced server spend at today's 3-4x inflated prices. By accelerating Spark workloads 10-20x, Speedata creates value by relieving the server price tsunami in a direct and measurable manner. Speedata.io
-
Bamiyan Gobets heeft dit gerepostBamiyan Gobets heeft dit gerepostWhy does a GPU running SQL feel like it's barely trying? Why are TPUs cloud-only? What does an LPU actually do that a GPU can't? The architectures are different because the workloads are different, and at production scale, those differences compound into real money. A breakdown of the five processors defining modern compute, and where each one fits at the link in the comments. #semiconductors #AI #APU #CPU #TPU #LPU #GPU #Nvidia #Google
-
Bamiyan Gobets heeft dit gerepostBamiyan Gobets heeft dit gerepostEvery Wednesday we host a live session on the Speedata APU: 20 minutes of Apache Spark SQL and AI data prep workloads executing on purpose-built silicon, plus an architecture walk-through. Today's session is at 1pm EST. To join, register at the link in comments. #AI #Analyticsacceleration #Apachespark
-
Bamiyan Gobets heeft dit gedeeldWe've just published the Speedata.io Live Stream Recording on our blog page: View here ❤️🔥 : https://lnkd.in/eudbFQ37Bamiyan Gobets heeft dit gedeeldWe hosted our webinar on the modern AI compute stack. The GPU-first model made sense when AI was experimental. In production, efficiency is the priority. Running the wrong workload on the wrong chip means overpaying in power, memory, and infrastructure costs. We broke down where each processor fits: 🟧 CPUs for orchestration and control logic 🟧 GPUs for model training and inference 🟧 TPUs and cloud ASICs for hyperscale AI 🟧 APUs for analytics-heavy data pipelines and AI data preparation 🟧 LPUs for low-latency inference decoding And why the agentic era is creating an entirely new pressure point on data infrastructure that most architectures weren't designed to handle. Full recap and recording now live on the Speedata blog. Link in comments. #AI #Semiconductor #GPU #ETL #Spark
-
Bamiyan Gobets heeft dit gerepostBamiyan Gobets heeft dit gerepostNVIDIA & AMD GPUs were designed for AI training and inferencing. Speedata.io APUs were designed from scratch for analytical data processing. APUs are purpose built for OLAP workloads including Apache Spark Velox Apache DataFusion Apache Doris Amazon Web Services (AWS) EMR/Athena Microsoft Azure Fabric Google DataProc
-
Bamiyan Gobets heeft dit gerepostBamiyan Gobets heeft dit gerepostAI agents can ask analytics questions. But can your infrastructure answer them fast enough? Agentic Analytics, executing advanced analytics queries from an LLM, is only useful if the answer comes back in seconds, not hours. In our recent webinar, we walked through how the agentic analytics pipeline works and where the bottleneck lives. Watch the full recording — link in the comments 👇 #AI #Spark #LLM #AgenticAnalytics
-
Bamiyan Gobets heeft dit gerepostBamiyan Gobets heeft dit gerepost$1T 2030 Data Center Compute TAM GPU/TPU NVIDIA AMD Google Amazon Web Services (AWS) CPU AMD Intel Amazon Web Services (AWS) APU Speedata.io
Bamiyan Gobets heeft commentaar gegeven op een bijdrage
2 d
Very nice summary.
Bamiyan Gobets heeft gereageerd op commentaar
1 w
Matt Hymers actually totally agree with you and appreciate your authenticity. It’s just hard to separate anymore. Great post!
-
Bamiyan Gobets vond dit interessantBamiyan Gobets vond dit interessantGoldman Sachs predicts AI agent token use to grow 24x by 2030. Enterprise agents will soon generating more SQL queries than humans. OpenAI and Anthropic are partnering with Databricks and Snowflake because the foundation for enterprise AI agents is structured enterprise data. Speedata.io APUs can deliver 1-2 orders of magnitude better performance, price-performance and energy efficiency for SQL analytics compared to CPU and GPU.
-
Bamiyan Gobets vond dit interessantBamiyan Gobets vond dit interessantWe are living through three simultaneous transformations. The most consequential technology change since the birth of humanity. The most significant world order transition since the Cold War. And the largest industrial infrastructure built out in history. But these are not separate stories, they are all part of a single story. Now the key question at the center of it is whether open societies, the ones that require consent, debate, democratic participation, can build fast enough and fairly enough to stay sovereign. The biggest risk is not that we fail technically, it is not that we hit a wall, it is not even that China or some other rival outpaces us in raw capability. I think the biggest risk is that Western democracies look at this moment, see what is required of them -- the energy build out, the industrial mobilization, the policy reforms, the investment and the political courage to drive frontier Intelligence into the hand of every citizen - and they fail to rise to the occasion.
-
Bamiyan Gobets heeft hierop gereageerdBamiyan Gobets heeft hierop gereageerdYour analytics platform was built for humans. That’s the problem. A human analyst asks one question. An AI agent asks one question…then fires off 6, 8, 12+ queries behind the scenes: Schema discovery. Metadata inspection. Candidate SQL. Validation. Follow-ups. Drill-downs. “Wait, check that again.” Now multiply that by every PM, operator, sales person, finance partner, and executive suddenly talking to data through agents. This is the future everyone is excited about. It is also the future that breaks your lakehouse/warehouse bill. Snowflake and Databricks were built for humans. But agentic analytics is a different workload. It is repetitive. It is query-heavy. It is full of joins, filters, aggregations, transformations, and validation loops. The issue is that agents turn every business user into a query-generating machine. This is why Speedata.io APUs matter. GPUs were built for AI model math. CPUs were built to be flexible. Speedata.io APUs are purpose built for SQL analytics. They accelerate the actual data operations agents depend on: scans, joins, aggregations, transformations, Spark SQL, ETL, data prep. In other words: Agents create the query explosion. APUs absorb the analytics explosion. This doesn’t mean ripping out Snowflake or Databricks. It means the next architecture will not be “one platform runs everything.” It will be: Governance in the platform. Storage in open formats. Routing at the query layer. Acceleration in purpose-built silicon. The winning data stack won’t ask, “How do we make one warehouse do everything?” It will ask: “What is the right execution path for this workload?” Some queries stay in Snowflake. Some run through Databricks. Some route to cheaper OLTP engines. And the heavy analytics/data-prep layer gets accelerated by Speedata.io APUs. That is the shift. The agent era will not be won by throwing more general-purpose compute at a workload that is becoming more specialized every week. It will be won by specialization. LLMs need GPUs. Analytics agents need Speedata.io APUs. The companies that figure this out early will make Agents feel faster. The companies that don’t will discover that Snowflake Databricks Microsoft Fabric gets prohibilitively expensive. The next bottleneck in AI is not just tokens. It is data infrastructure. And the next unlock is not another dashboard. It is analytics at the speed of silicon. #AI #DataEngineering #Analytics #Databricks #Snowflake #APU #AgenticAI #DataInfrastructure
-
Bamiyan Gobets heeft hierop gereageerdBamiyan Gobets heeft hierop gereageerdTwo projects. Two Gold awards. A memorable night at the Dutch Interactive Awards. 🏆 Gold in Gen AI & Agents with Fidacta 🏆 Gold in Tools (Interactive Applications) with the AI Product Locator for Maxeda DIY Group, Praxis and Brico A huge thank you to our clients, partners, and colleagues who helped make this possible. We’re honored to see our work recognised and look forward to building what’s next. #DIAAwards #DIA #GenAI #ProductInnovation #AIAgents
-
Bamiyan Gobets vond dit interessantBamiyan Gobets vond dit interessantSQL expertise has 5 levels. - Level 1: Basics SELECT, FROM, WHERE, GROUP BY, HAVING, LIMIT. These are the SQL keywords you use in every query. With them, you can: → Filter data → Sort results → Build basic reports → Aggregate a bunch of records - Level 2: Joins This is where SQL starts becoming powerful. Most of the time, you’ll use: → INNER JOIN → LEFT JOIN And much less often: → RIGHT JOIN → CROSS JOIN → FULL OUTER JOIN If you understand joins well, you understand SQL well. - Level 3: Window Functions This is where SQL becomes a serious skill. You need to understand: → PARTITION BY: split the window → ORDER BY: order the window And the difference between: → RANK → DENSE_RANK → ROW_NUMBER Window functions are one of the biggest jumps from beginner to intermediate SQL. - Level 4: The Architect You can’t just query the data. You also need to understand how the structure is built. That means knowing DDL: → CREATE → ALTER → DROP And also understanding transactions: → COMMIT → ROLLBACK Because SQL is not only about reading data. It’s also about designing and managing it correctly. - Level 5: The Optimizer This is the superior level. You don’t just write SQL. You understand: → Indexes → Partitions → Table scans → Query performance → How the database actually executes your query 2 queries can return the same result. But one can run in 2 seconds. And the other can destroy your warehouse. --- 📤 Send this to a friend to know their SQL level ♻️ Repost this if you found it useful. 👉🏻 Follow José for more about Data and AI
-
Bamiyan Gobets heeft hierop gereageerdBamiyan Gobets heeft hierop gereageerd💡 The Six Stages of a Data Analysis Project ⭕ 1. Data Requirement Every data project starts with a pause. Before opening a spreadsheet or writing a query, you ask yourself what you’re actually trying to figure out. What decision is on the line? What question needs an answer? If this part isn’t clear, everything that follows becomes busy work dressed up as analysis. ⭕ 2. Data Collection Once the question is clear, you go looking for information that might help answer it. You pull data from databases, files, logs, or APIs, knowing full well that some of it will be messy or irrelevant. That’s fine. At this stage, you’re collecting, not judging. You just want enough material to work with. ⭕ 3. Data Processing Now you try to make sense of what you’ve gathered. Raw data rarely tells a story on its own, so you start organizing it. Tables get joined, columns get renamed, numbers get grouped. You’re not looking for insights yet. You’re just turning chaos into something readable. ⭕ 4. Data Cleaning This is where you slow down and start questioning everything. You notice missing values that shouldn’t be missing and numbers that don’t make sense. Some records repeat themselves. Others feel off. You fix what’s broken because you know that clean data isn’t about perfection, it’s about trust. ⭕ 5. Data Analysis With clean, structured data in front of you, patterns begin to show up. Trends move in directions you didn’t expect. Comparisons suddenly make sense. This is the moment where curiosity kicks in, and you start asking better questions than the ones you started with. ⭕ 6. Data Interpretation You step away from the numbers and think about what they actually mean. Not for you, but for the people who need to act on them. You translate insights into plain language and connect them back to the original question. This is where analysis turns into decisions. ⚡ And the truth is, you rarely move through these steps in a straight line. You’ll circle back, change your question, clean the data again, or collect something new. That’s not a failure. That’s real data work.
Ervaring en opleidingen
-
Speedata.io
Volledige ervaring van Bamiyan weergeven
Bekijk onder meer de functietitel en de aanstellingsduur.
of
Talen
-
English
-
-
Dutch
-