Who Created Deepseek AI?

Who’s behind the AI revolution transforming technology?

This blog post reveals the creators of DeepSeek AI, a company making waves with its innovative and efficient AI models.

We’ll explore the founders, history, and groundbreaking tech that has shaken Wall Street and ignited a global discussion about AI’s future.

As the AI field grows, many are exploring different options, and if you’re curious about a DeepSeek AI alternative, this post will give you valuable background on a key player in the industry.

The Founder: Liang Wenfeng

DeepSeek AI was founded by Liang Wenfeng, a remarkable entrepreneur blending financial expertise with a passion for artificial intelligence.

Liang’s journey started at Zhejiang University, where he first explored automated stock trading before focusing on AI’s exciting potential. In 2015, he co-founded High-Flyer, a quantitative hedge fund pioneering AI in trading strategies.

This experience was invaluable, setting the stage for DeepSeek AI’s future success.  

DeepSeek AI’s Beginnings

DeepSeek AI’s origins lie within High-Flyer’s artificial general intelligence lab, established in April 2023.

Initially driven by Liang’s scientific curiosity and a desire to push AI’s limits, DeepSeek AI officially came into existence on July 17, 2023.

While venture capital firms hesitated to invest, Liang’s unwavering belief in DeepSeek AI’s potential, along with High-Flyer’s financial backing, allowed the company to thrive.

Interestingly, DeepSeek AI’s emergence coincided with the United States’ restrictions on AI chip exports to China.

Despite these limitations, DeepSeek AI acquired a substantial stockpile of Nvidia GPUs, including the H800s, designed to circumvent the initial export controls.

Some reports even suggest DeepSeek AI might have access to the more advanced H100s.

This access to significant computing resources was crucial in DeepSeek AI’s ability to develop highly efficient AI models.

Furthermore, the timing of DeepSeek AI’s R1 model release, coinciding with the US presidential inauguration, suggests a strategic move to challenge the US’s perceived dominance in AI.

This bold move immediately captured the global tech community’s attention and sent shockwaves through Wall Street, impacting the stock prices of major tech companies like Nvidia and ASML.  

The Vision and Mission

DeepSeek AI’s mission is to make AI technology accessible to everyone, from businesses and researchers to individuals.

The company envisions a future where AI empowers various industries, enhances human capabilities, and provides solutions to complex global challenges.

This commitment to accessibility is evident in DeepSeek AI’s open-source approach, which allows for the free use, modification, and examination of its AI models.  

The Team

DeepSeek AI’s success is largely due to its talented team of AI engineers and researchers.

Unlike many other AI companies that prioritize experienced engineers and product development, DeepSeek AI focuses on technical abilities and has built a team primarily composed of recent graduates from top Chinese universities. T

This unique approach fosters a dynamic and innovative environment where young talent can thrive and contribute to cutting-edge research.

DeepSeek AI is known for offering competitive salaries, rivalling those of industry giants like ByteDance, to attract and retain these bright minds.  

Products and Services

DeepSeek AI has developed an impressive array of AI models, each with unique capabilities and applications:  

• DeepSeek-V3: Launched in December 2024, this powerful large language model is known for its exceptional performance and efficiency. DeepSeek-V3 utilizes innovative techniques like sparsity, where only a small fraction of the model’s parameters are used for any given input, and memory compression, allowing for efficient storage and access of information. These techniques contribute to the model’s impressive performance with reduced computational resources.  
• DeepSeek-R1: Released in January 2025, this reasoning model is designed to tackle complex problems by employing enhanced reasoning capabilities. It excels in tasks that require deep analysis and strategic thinking.  
• DeepSeek-R1-Zero and DeepSeek-R1-Distill: These variations of the R1 model offer different parameter sizes, providing flexibility and accessibility for various applications.  
• Janus-Pro-7B: This vision-based model, launched in January 2025, expands DeepSeek AI’s capabilities beyond language processing, demonstrating the company’s commitment to exploring diverse AI applications.  

Product/Service	Description	Key Features	Launch Date
DeepSeek-V3	Large language model	Strong performance, efficiency, sparsity, memory compression	December 2024
DeepSeek-R1	Reasoning model	Enhanced reasoning capabilities, tackles complex problems	January 2025
DeepSeek-R1-Zero & DeepSeek-R1-Distill	Variations of R1 model	Different parameter sizes, flexibility for various applications	January 2025
Janus-Pro-7B	Vision-based model	Expands capabilities beyond language processing	January 2025

DeepSeek AI’s commitment to open-source principles is further exemplified by the release of its models under the MIT License.

This allows anyone to access, use, and modify the models, fostering collaboration and knowledge sharing within the AI community.

These models are readily available on platforms like Amazon Bedrock and Azure AI Foundry, making them accessible to a wider audience and increasing their potential for widespread use.  

Cost-Effectiveness and Efficiency

One of DeepSeek AI’s most notable achievements is its ability to develop high-performing AI models at a fraction of the cost of its competitors.

DeepSeek-V3, for instance, was reportedly trained for approximately $5.58 million, significantly less than the $100 million spent on developing OpenAI’s GPT-4.

This cost-effectiveness stems from DeepSeek AI’s innovative approach to AI development, which prioritizes efficiency and resource optimization.  

Security and Ethical Considerations

While DeepSeek AI has made remarkable strides in AI technology, it’s important to acknowledge the security vulnerabilities and ethical considerations surrounding its models.

Research has shown that DeepSeek’s models can be susceptible to jailbreaking, generating harmful content, and producing hallucinations.

These findings highlight the need for ongoing research and development to address these challenges and ensure the responsible use of DeepSeek AI’s technology.  

DeepSeek’s Global Impact

DeepSeek AI’s rise has not only disrupted the AI industry but also sparked a broader conversation about the global AI landscape.

The company’s ability to develop cutting-edge AI with limited resources challenges the US’s dominance in the field and highlights the growing competitiveness of Chinese AI companies.

This has led to concerns among US tech investors about the potential shift in the balance of power in the AI world.

Moreover, DeepSeek AI’s data privacy practices, with data storage on servers located in China, have raised concerns among some experts.  

Challenging the AI Status Quo

DeepSeek AI’s success has challenged several long-held assumptions in the AI industry.

Firstly, it has demonstrated that massive computing power may not be essential for developing state-of-the-art AI.

By focusing on efficiency and innovative techniques, DeepSeek AI has achieved remarkable results with fewer resources.

Secondly, DeepSeek AI’s team composition and research-oriented approach differentiate it from other AI companies that prioritize product development and experienced engineers.

This unique strategy has allowed DeepSeek AI to focus on pushing the boundaries of AI research and development.  

Conclusion

DeepSeek AI’s journey is a testament to the vision and dedication of its founder, Liang Wenfeng, and the talented team he has assembled.

By prioritizing research, efficiency, and accessibility, DeepSeek AI has not only developed groundbreaking technology but also challenged the status quo in the AI industry.

DeepSeek AI’s open-source approach and cost-effective models have the potential to democratize AI development and reduce the dominance of large tech companies.

With its commitment to innovation, efficiency, and open-source principles, DeepSeek AI is not just a rising star in the AI landscape, but a potential catalyst for a new era of accessible and democratized artificial intelligence.