Suggestions
Yonatan Geifman
Co-Founder & CEO at Deci (We are hiring!)
Yonatan Geifman is a prominent figure in the field of artificial intelligence and deep learning. He is the Co-Founder and CEO of Deci, a company that was recently acquired by NVIDIA.2 Geifman's work focuses on building AI software and foundation models, and he is now associated with NVIDIA following the acquisition.1
Expertise and Contributions
Geifman is known for his expertise in:
- AI and Deep Learning: He frequently shares insights and research on advanced prompting techniques for AI.1
- Model Optimization: Geifman specializes in improving and scaling deep learning models' inference performance.4
- Edge AI: His work involves applying AI to build better AI solutions, both at the edge and in the cloud.3
Professional Activities
Yonatan Geifman is actively involved in the AI community:
- He conducts live sessions on model inference performance optimization.4
- Geifman participates in discussions about boosting inference performance for AI models.3
- As the founder of Deci, he has been instrumental in developing AI technologies that have caught the attention of major players like NVIDIA.2
Geifman's LinkedIn presence (username: yonatan-geifman-8a449a3) serves as a platform where he shares valuable insights and updates about advancements in AI and deep learning technologies.14
Highlights
There's a great opportunity to join our team and work in one of the most interesting areas of LLMs inference software. Contact me for more details.
I’m excited to share that Deci AI is collaborating with Qualcomm to streamline deployment of generative AI and LLMs on Qualcomm hardware. With the release of DeciCoder-6B and DeciDiffusion 2.0, both models are engineered for accurate and efficient performance at scale, on the Qualcomm Cloud AI 100, we’re making AI more accessible and cost-effective, and economically viable for a wider range of applications. 𝐃𝐞𝐜𝐢𝐂𝐨𝐝𝐞𝐫-6𝐁 is a multi-language code LLM that supports 8 languages, outperforms CodeGen 2.5 7B, CodeLlama 7B, and StarCoder 7B, with 3 points advantage in Python over StarCoderBase 15.5B. It also achieves 19x higher throughput compared to CodeGen 2.5 7B. 𝐃𝐞𝐜𝐢𝐃𝐢𝐟𝐟𝐮𝐬𝐢𝐨𝐧 2.0 is text-to-image diffusion model that outperforms Stable Diffusion 1.5, enabling 2.6x faster, cheaper image generation with image quality on par with Stable Diffusion v2.1. It generates images in under 1 second for a fraction of the cost. Generated by Deci’s AutoNAC, a NAS-based engine that is hardware-aware, both models are optimized to leverage the full computational power of Qualcomm’s Cloud AI 100. We can’t wait for everyone – from enterprises to the community – to push the boundaries of what's possible in AI efficiency and performance together with us at Deci, and our partner, Qualcomm. 𝐃𝐨𝐰𝐧𝐥𝐨𝐚𝐝 𝐚𝐧𝐝 𝐥𝐢𝐤𝐞 𝐃𝐞𝐜𝐢𝐂𝐨𝐝𝐞𝐫 6𝐁 𝐚𝐧𝐝 𝐃𝐞𝐜𝐢𝐃𝐢𝐟𝐟𝐮𝐬𝐢𝐨𝐧 2.0 𝐨𝐧 𝐇𝐮𝐠𝐠𝐢𝐧𝐠 𝐅𝐚𝐜𝐞 >
🤗 DeciCoder-6B Model Card: https://lnkd.in/dDQVyerV 🤗 DeciDiffusion 2.0 Model Card: https://lnkd.in/d-BuBheg 📘Technical Blog : https://hubs.ly/Q02gtXS70 📔DeciCoder-6B Notebook: https://lnkd.in/dYaqcPZs
#decicoder #decidiffusion #generativeai #largelanguagemodels #llm
