I’m currently making a career change from backend web development (PHP/MySQL stack) to artificial intelligence engineering within my company. Been working on this transition for about 8 weeks now.
Right now I’m focusing on connecting large language models using tools like LangChain and LangGraph, plus monitoring these systems with LangSmith. I’m also building RAG systems with vector databases like ChromaDB to handle company-specific data and reduce AI hallucinations.
My learning roadmap includes mastering LangSmith for AI agents and function calling, then moving into model fine-tuning and eventually multimodal AI applications involving image processing.
Currently using Django and FastAPI for most projects, but it feels like I’m still doing regular web development just with AI API calls mixed in.
My goal is landing a dedicated AI engineering position in the next few months. For those already working in AI roles - what does your typical workday involve? Are you dealing with similar technologies or is the field completely different from what I’m experiencing?
I’m new to this space and learning everything from scratch, so any guidance on which skills to prioritize would be amazing. Also open to resource recommendations!
the hardest part isn’t the tech stack - it’s shifting from deterministic to probabilistic thinking. in php, things either work or they dont. with ai, outputs are always wrong in different ways. focus on understanding when and why your rag system fails. get comfy with uncertainty metrics and confidence scores - that’s what separates ai engineering from regular backend work.
Your timeline works, but focus way more on evaluation. Most companies can’t measure their AI systems properly - that’s the real problem, not building them.
I spend 60% of my time monitoring and improving existing workflows. The rest splits between new features and production edge cases.
Backend to AI isn’t a huge jump. You’re still solving data problems with different tools. Vector databases are just regular databases once you understand embeddings.
Don’t bother with model fine-tuning yet unless you’ve got a specific need. Better prompts and retrieval solve most business problems - not custom models. Companies want people who ship reliable AI features quickly.
What really helped my transition was building evaluation datasets early. Create test cases for RAG systems and measure retrieval quality. That’s what separates junior from senior AI engineers.
Learn async programming if you haven’t. AI workflows need tons of concurrent API calls, and you’ll constantly hit rate limits without proper queuing.
Your PHP background actually helps. You know how to build systems that handle failures well, which matters when AI APIs crash or return junk responses.
You’re spot on - what you’re experiencing is exactly what AI engineering looks like day-to-day. That ‘regular web development with AI API calls’ feeling? That’s because most production AI work IS building pipelines, handling data flows, and creating solid interfaces around AI components. My typical days involve debugging model outputs, optimizing retrieval systems, watching performance metrics, and tons of regular software engineering. The RAG work you’re doing with ChromaDB is perfect since most commercial AI apps lean heavily on retrieval systems instead of custom trained models. I’d push you to go deeper into prompt engineering and evaluation frameworks beyond LangSmith. Learning to systematically test and improve AI outputs becomes critical when you’re running production systems. Also get familiar with MLOps tools - deploying and maintaining AI systems is different from traditional web apps. Your Django/FastAPI background is actually gold here. Lots of AI teams need people who know how to build scalable web services around AI components. You’re building the right technical foundation for today’s job market.