deepsense.ai

22/10/2024

In the fast-paced world of e-commerce and beyond, visual search transforms how consumers discover and purchase products. This lightning session showcases deepsense.ai’s and Anyscale's cutting-edge fashion image retrieval system—a powerful solution that combines the strengths of Anyscale Ray and Pinecone.

At Ray Summit 2024, Artur Zygadło explored the architecture of this intuitive and scalable application, which leverages CLIP embeddings and vector similarity search to integrate text-to-image and image-to-image queries seamlessly. He detailed how Ray Serve facilitates autoscaling for optimal performance and demonstrated advanced parallelization techniques using Ray Data.

This talk provides a practical blueprint for developers aiming to build high-performance, user-friendly cross-modal search applications, not only in retail but in other sectors as well.

Watch the full session on YouTube: https://www.youtube.com/watch?v=q4Na3q_gQfk

18/09/2024

Are your LLM-based applications ready for the constantly changing world of upgrades?

This article breaks down how LLMOps can streamline deployment, optimize performance, and ensure your AI investments deliver consistent value. Whether you're dealing with data drift, cost overruns, or integration challenges, LLMOps offers the solutions you need to operationalize your LLMs successfully.

Curious about how to make your LLMs more robust and reliable?
👉 https://deepsense.ai/operationalizing-large-language-models-how-llmops-can-help-your-llm-based-applications-succeed/?utm_source=facebook&utm_medium=post_190924&utm_campaign=LLM

04/09/2024

Join us as Bill Salak shares his journey in leading the development of an AI-powered platform, Brainly, that serves millions daily. In our interview, Bill explores the universal technological challenges and breakthroughs in building and scaling AI products for global users. And yes, this is especially crucial because it’s shaping the future of education!

💡 Key topics:

1. Implementation

The AI implementation process at scale, from ideation to deployment

2. AI reliability

Managing AI reliability and ensuring accurate, data-driven outcomes

3. MLOps

The role of MLOps and standardized tools in accelerating AI growth

4. AI development

Staying ahead of AI advancements and adapting to fast-evolving technologies

5. AI impact on users

Strategies for measuring AI's impact on user experience and business outcomes

This interview is packed with actionable insights for CTOs and AI leaders looking to build robust, scalable AI systems that deliver real-world value.

📺 Watch our interview on YouTube: https://www.youtube.com/watch?v=TPJTkJnfmgE

PS And kudos to Alicja Stecz and Jan Kanty Milczek for hosting the event and asking those crucial questions!

28/08/2024

🚀 Great news for 3D gaussian splatting developers interested in custom modifications on forward/backward kernels with a minimal starting point, and teams ready to fully embrace open-source solutions in their workflows!

If you are utilizing 3D reconstructions with Gaussian splatting in your project, we’ve developed a solution tailored just for you!

What are the features?
🔹 Minimalistic design
Our rasterizer features a streamlined codebase, focusing solely on the essence to ensure simplicity and efficiency.

🔹 Easy integration
Seamlessly switch to our implementation using the same interfaces as the vanilla version. This makes integration and testing straightforward, allowing you to maintain full functionality with ease.

🔹 Optimized performance
Powered by our own fast CUDA kernels for both forward and backward operations, our rasterizer delivers exceptional speed without compromising on performance.

🔹 End-to-end environment example
Whether you're starting a project from scratch or enhancing an existing setup, check out our integration with Gaussian Splatting Lightning in this example: https://github.com/deepsense-ai/ds-splat?tab=readme-ov-file

🔹 Open-source
Transition entirely to open-source with our commercially friendly Apache 2.0 licensed code, ensuring flexibility and freedom for your projects.

🔗 Dive into our code on Github: https://github.com/deepsense-ai/ds-splat

Give it a try, and share your feedback with us!

22/08/2024

🐻 Polars may be vulnerable in the wild, but not in the world! The Polars DataFrame library, a fresh competitor, is rapidly gaining popularity and challenging the old and infirm king, Pandas. 🐼

Built with Rust and Python, Polars has quickly proven itself with powerful features like out-of-core processing, parallelism, and zero-copy data exchange. With its 1.0 release in July 2024, Polars offers top-tier performance and a developer-friendly API reminiscent of PySpark, a comprehensive set of nullable data types, and a wealth of I/O connectors.

Expect a unique webinar experience that matches the cutting-edge content—as you can see from the screenshot. It's a good thing we have Jakub Cierocki as our Senior Machine Learning Engineer, not a graphic designer. 🤡

Don’t miss this chance to explore why Polars might just be the future of data manipulation in Python:
👉 https://www.youtube.com/watch?v=8oMx-DD0tfk&utm_source=facebook&utm_medium=social&utm_campaign=deeptalk

14/08/2024

❓ Struggling to effectively analyze and visualize new datasets in your AI projects? You’re not alone. Discover what we've learned from working with OpenAI's Data Analyst and Microsoft's LIDA, and the insights we've gained through Exploratory Data Analysis (EDA): https://www.youtube.com/watch?v=DfJQuzIr65w&utm_source=facebook&utm_medium=social&utm_campaign=deeptalk

In our latest webinar, our ML Engineer Michał Grochecki dives deep into EDA, revealing strategies to:
🔍 Discover patterns
⚠️ Spot anomalies
🧠 Form and test hypotheses
✅ Check underlying assumptions

Plus, learn how cutting-edge tools like OpenAI's Data Analyst and Microsoft’s LIDA can supercharge your workflow with LLM-powered visualization, recommendations, and automated insights.

Don’t miss out on mastering your data with minimal effort—watch the full session now on YouTube: https://www.youtube.com/watch?v=DfJQuzIr65w&utm_source=facebook&utm_medium=social&utm_campaign=deeptalk

Speaker: Michał GrocheckiMachine Learning EngineerDo you get overwhelmed when dealing with a new dataset? Are you struggling to find ways to analyze and visu...

07/08/2024

Are your LLMs acting up with hallucinations, data leaks, or off-target responses?

Join our Senior ML Engineer, Alan Konarski, in an insightful talk about fixing these common headaches with a few steps:

🌟 Why monitoring matters?
📊 Which numbers really matter when it comes to making your LLMs sing?
🛠️ Get the scoop on the best monitoring stack to keep your LLM applications in check

Check out the full video and let us know what you think.

🔗 Watch now on YouTube: https://www.youtube.com/watch?v=oWAgaMjo3Do&utm_source=facebook&utm_medium=social&utm_campaign=post

31/07/2024

We had the privilege of sitting down with Burkhard Boeckem, CTO of Hexagon AB, for an insightful and engaging discussion. Here's a sneak peek into the key topics of the conversation:

💠 STRATEGY
AI development strategy and focus of enterprises in terms of AI.

Hexagon’s strategy in balancing organic innovation with strategic acquisitions to enhance their AI capabilities.

💠 PRODUCT DESIGN
How AI enhances user-centric product design.

💠 DATA AND ANALYTICS
How Hexagon leverages AI to process vast amounts of data from industrial sensors, enabling real-time analysis and smart digital realities.

💠 AI DEVELOPMENT STRUCTURE
Insights into Hexagon’s internal AI development structure, comparing the roles of business divisions and a central AI hub.

The potential changes AI will bring to the workforce and the importance of preparing employees for new skill requirements.

💠 FUTURE INNOVATIONS
AI innovations on the horizon, especially those poised to transform Hexagon's offerings and expand the industrial metaverse.

Don't miss out on this opportunity to explore the future of AI with one of the industry's leading voices. Watch the full interview hosted by Michal Tadeusiak and Jan Kanty Milczek for more in-depth insights and inspiring discussions!

🔗 https://www.youtube.com/watch?v=KyFekiNQszk&utm_source=facebook&utm_medium=social&utm_campaign=video

19/07/2024

🚚 🚛 🚒 Struggling with limited computing power in your data science and machine learning projects? We reveal practical strategies to optimize your resources and enhance productivity.

Leverage tools like SLURM, network FS, and LMOD for a smoother workflow. Don’t let limited resources slow you down!

📖 Read the full article: https://deepsense.ai/optimizing-computational-resources-for-machine-learning-and-data-science-projects-a-practical-approach/?utm_source=facebook&utm_medium=social&utm_campaign=blog

16/07/2024

🤔 How Can AI Revolutionize Ada Development? Discover how our AI Copilot, developed with AdaCore, can transform software developers’ productivity. This proof-of-concept tool offers intelligent code completions, automates repetitive tasks, and significantly accelerates development.

Curious? Dive into the entire case study and watch the demo to see the future of Ada development!
👉 https://deepsense.ai/ai-copilots-impact-on-productivity-in-revolutionizing-ada-language-development/?utm_source=facebook&utm_medium=social&utm_campaign=case