Series Catalog
1.Programmer Transformation AI: Industry Analysis
2. Programmer Transformation AI: Transformation Plan
3. Programmers transform into AI: implementation practice
4. Programmers transform into AI: Looking to the future
1. Background analysis
Entering 2025, AI has grown explosively and has entered the stage of actual commercial monetization. It is said that "when standing on the wind vent, pigs will fly, but the wind vent disappears.Pigs will fall down or even die”. But the blogger sawAI is already a trend! ! ! ! Not just the wind gust. Similar to the strong rise of Java language in enterprise-level application programming in those days, SpringCloud framework in the framework is unstoppable, andAI will explode and grow for 5 years, and will operate normally for 5-10 years., this willalmostyesThe only opportunity for ordinary programmers to overtake (already owned high-level programmers in large factories), it is also a magic medicine to prolong your career (simply put, it is to solve the crisis of 35 years old...).
The blogger is relatively conservative, but is currently forced to be an agent (the concept is more advanced than the existing ordinary meaning), but found thatThe concept is vague, the land is difficult, and the theory and practice are lacking... It is trapped in a quagmire. I will review this article and record my own transformation path (the process of traditional Java development engineer/architectural transformation into AI and embracing AI). I hope tooFor programmers who are still confused, one way to improve. ---------To put it bluntly, Jack Ma has made a comeback, are you still waiting? ? ? AI is unstoppable. If AI cannot be applied within 2-3 years, it may be directly eliminated by the industry. You may not believe it, but you can wait and see.
2. Industry overview
2.1 Industry Analysis
There are many things that can be done around AI. We do industry analysis around traditional programmers/architects/technical managers, what transformation AI can do:
algorithm
Pure algorithm, generally a highly educated algorithm engineer with extremely strong algorithm capabilities, usually onlyLarge companies, research institutionsJust want it.
application
1) AI framework development---from 0 to 1
Framework/platform AI developers: Develop general AI frameworks, such as the first open source deepseek, which is a quantitative domestic magic square, and the first closed source chatGPT, which is a subsidiary of openAI. Of course, there are also: Alibaba Tongyi Qianwen and Baidu Wenxin Yiyan and other complete sets of platform-level self-developed AI frameworks. This type of company is generallyLarge factorylevel.
2) Training fine-tuning, secondary development---from 1 to 1.1
In-depth developer: Companies with such developers are generally - companies with their own detailed needs and high-value data in vertical industries. Companies with the ability or expect to develop AI frameworks secondary to better adapt to their business. At least such companies areUnicorn/large traditional digital transformation company。
3) Research and development of AI frameworks, application development--from 1 to N
Application Developer: For example: The solution directly connects the three-party AI framework to realize its own product functions.Ordinary technology companiesThat's it, for example, international translation access to machine translation gpt-4o.
4) Non-R&D, pure AI tool users---Derivatives
Tool User:Training around AI skills (various traffic IP/self-media teachers); content creators (text, pictures, videos, etc.) that AIGC uses AI tools, can beAny company/individual；
In summary,There are many positions and suitable for the transformation of older ordinary programmers, AI-related positions: 2 and 3.. That is, AI-related second-open/application developer/architect. If you have the opportunity to do the second category directly: the position of in-depth developers is better. Of course, if you lack the ability temporarily, you can first enter the three-party AI framework and directly dock the application development. Because when your business grows and grows, you are likely to find the lack of tools. After that, you have to upgrade and fine-tune yourself to better adapt to your business needs scenarios. Anyway,If you have the opportunity, you can get in first and then change it. This idea is correct.！
Moreover, whether it is in-depth developers (training fine-tuning) or simple application development, there are currently many positions in boss direct recruitment, and there are comprehensive explosive growth. The market is extremely lacking, and the salary is about 40-80K, which is far higher than Ordinary programmer.
2.2 Technology Stack
We narrowed our scope and focused on the above positions that are easy to transform. That is, AI training fine-tuning, secondary development and AI application development, sort out relevant positions, and finally sort out the corresponding technology stack.
2.2.1 Common AI Job Technology Stack
- Big model trainers: GPT, ChatGLM, LLaMA, MM-LLMs. Basic algorithm principles, Fine-tuning strategy, Prompt engineering, vector database. Common distributed training solutions, Megatron, DeepSpeed, etc.
- Deep Learning Framework: Proficient in TensorFlow/PyTorch.
- AI Engineer: Big model construction + training. Build the capability of large-scale RAG/Agent.
- Data engineering: traditional big data technology stacks such as big data processing, search algorithms, and machine learning (hadoop, spark, flink, etc.)
Programming language: a kind of proficient in python and proficient in go/c++/java.
2.2.2 Other related technology stacks
- AI low code: low-code platform low-code, etc. It turns out that it is still possible to automatically generate code in a low-code platform in a simple business scenario.
- AIOT: AI is combined with the Internet of Things, which is a trend. AIOT is unstoppable.
- Edge AI: One of the overall solutions for the integration of cloud edge and edge in smart devices. The edge AI part will directly sink part of the logic from the cloud to the edge and end to execute.