Sign In

Raman SHRIVASTAVA

AI Engineer & Leader | LLMs, RAGs, AI Agents | 40 under 40 Data Scientist, AIM 2019

Raman Shrivastava is currently the AI Lead at INFINIT, a position he has held for several years. With over 13 years of professional experience, he has dedicated the last 10 years specifically to the field of Artificial Intelligence. His expertise encompasses various aspects of AI, including machine learning and data processing technologies.

Professional Background

  • Current Role: AI Lead at INFINIT
  • Location: Paris, France
  • Previous Experience: He has worked in various roles at companies such as CustomerGlu, Marax AI, and Fitsome Inc., which have contributed to his extensive knowledge in AI applications.13

Contributions and Focus Areas

Raman has been involved in innovative projects aimed at enhancing AI capabilities. For instance, he recently highlighted a project called Crawl4AI, which simplifies web crawling and data extraction processes, making them more efficient for large language models (LLMs).2 His work emphasizes the integration of AI technologies into practical applications that can drive business value.

Educational Background

Raman's educational qualifications include training from Udacity, which is known for its focus on technology and AI-related courses.1

Overall, Raman Shrivastava's role at INFINIT positions him as a significant contributor to advancements in AI, particularly in how these technologies can be leveraged for improved data handling and decision-making processes.

Highlights

Aug 4 · linkedin

Github models: https://lnkd.in/gDTprGjF

key points:

  • Models available: Llama 3.1, GPT-4o, GPT-4o mini, Phi 3, Mistral Large 2 etc.
  • Playground(low/no code): access each model via a built-in playground that lets you test different prompts and model parameters in GitHub.
  • Codespaces & VsCode integration: quick path to bring the models to developer environment in Codespaces and VS Code.
  • Azure AI integration: easy integration for enterprise production grade deployments
Jul 31 · linkedin

Key points from the Nvidia H100 whitepaper

• H100 Tensor Cores: 6x AI performance boost vs A100  ○ New FP8 precision  ○ Enhanced large language model processing

• FP8 precision advantages:  ○ 2x throughput of FP16/BF16  ○ 50% smaller memory footprint  ○ E4M3 and E5M2 formats

• Transformer Engine:  ○ Dynamic FP8/FP16 precision switching  ○ 9x faster AI training  ○ 30x faster inference on large models

• First GPU with hardware-based Confidential Computing  ○ Trusted Execution Environments (TEEs)  ○ Enhanced data/code protection

• NVLink Network:  ○ Connects up to 256 GPUs  ○ 57.6 TB/sec all-to-all bandwidth

• DPX instructions:  ○ 7x speedup for dynamic programming algorithms  ○ Major boost for genomics processing

• Asynchronous execution improvements:  ○ Tensor Memory Accelerator  ○ Asynchronous Transaction Barrier

• HBM3 memory on H100 SXM5:  ○ 3 TB/sec bandwidth  ○ 2x increase over A100's HBM2

• Second-gen MIG technology:  ○ 3x more compute per GPU Instance  ○ 2x more memory bandwidth vs A100

• DGX H100 system:  ○ 32 petaFLOPS of AI performance  ○ Scalable to DGX SuperPOD configurations

Related Questions

What are Raman Shrivastava's main responsibilities as AI Lead at INFINIT?
What projects has Raman Shrivastava led at INFINIT?
How did Raman Shrivastava get started in AI?
What are some notable achievements of Raman Shrivastava in the AI field?
What is Raman Shrivastava's educational background?
Raman SHRIVASTAVA
Raman SHRIVASTAVA, photo 1
Raman SHRIVASTAVA, photo 2
Add to my network

Location

Paris, Île-de-France, France