Navigating the GenAI hype within a large software dev company, I have outlined the following themes, distinct in nature (of work) yet easily confused:
1. Integrating GenAI into Software
E.g. building a chatbot relying on OpenAI chat completions APIs, integrating RAG and vector DBs, building cloud-native apps with services from Vertex AI, etc.
This part involves most of the devs and conceptually is no different from utilizing a 3rd party lib or cloud service API requiring a "black-box" understanding of the service: understanding inputs and output, not caring much about the internals.
2. Building GenAI
Fine-tuning models, deploying and operating specialized models, etc.
We are talking about data science, engineering, and MLOps here. White box and a deep understanding of the under-the-hood processes are required.
3. Building software with GenAI
Boosting engineering productivity, utilizing CoPilot, asking ChatGPT to generate acceptance criteria, etc.
The last part concerns everyone involved in the development, BAs, QAs, Devs, Managers - the aim is to make the whole SDLC process more efficient and effective, doing more with less.
Hope the paradigm helps:)
Top comments (0)