The Shift: Moving into AI Engineering
Long time no post! As you can imagine with this post title, I’ve been pretty focused on my career growth and trying to negotiate a specialization as an AI engineer. I’m sure my blog gives you an idea of my interests in AI, but I always desired to make it a professional job rather than a side hobby. To be transparent, my current company leadership was rightfully hesitant to adopt AI, as some folks worried about the security of it all. For almost a year I have been diligently doing research write ups to share with my boss and company, as well as even doing demos of what’s possible, which definitely sold it to the stakeholders! After much patience, they were thrilled with what I showed is possible, and now we are chasing AI initiatives, along with me moving into the role of “AI Development Specialist” (I was told it was difficult to create a new position title and this is what they decided on… but the new role description is basically an AI Engineer)
- Previous Role: Software Developer
- New Role: AI Development Specialist (AI Engineer)
- Key Shift: Moving from building traditional CRUD apps/services to designing LLM workflows, RAG pipelines, and fine-tuning models, as well as AI governance responsibilities.
Why the Transition Made Sense
As mentioned, I was already heavily advocating for AI in the company. We have a bunch of useful data, so why not put AI to work? My demos were not just snazzy, and they did not simply solve a problem. They actually addressed a concern the company had- data privacy and safety. If our data was leaked, it could put the company under VERY quickly.
- I always wanted to be an AI engineer, but I needed to envision how I can apply that desire and put it into action in my current company
- The demo project used a very niche tool called ML.net which allowed us to keep the entire process in our data center, meaning solving a logistics and safety problem in one demo.
- Personally, my degree and previous roles were not pure AI focused. What helped though was learning and using AI as a hobby. I would take courses in my free time, and build RAG models or agentic systems. All these things gave me the confidence to advocate for myself and confidently show management what AI is capable of.
What Changes Day-to-Day
It’s early to say since I just got the role last month, but based on discussions and the new job description, I have some notion of what it entails. Previously, I was only reporting to my manager and all my responsibilities came from him. I was never “owning” any major initiatives. Now, I am going to be expected to step up, which means owning the project (this is more than just “owning” the code), meet with stakeholders compared to the past where I was only meeting with my manager, and mediating/managing contractors.
Traditional Software Development vs. AI Engineering
| Aspect | Software Developer | AI Development Specialist |
|---|---|---|
| Primary Focus | Deterministic logic, API architecture, UI/UX | Probabilistic systems, LLM orchestration, AI governance |
| Tooling | Standard frameworks, databases, CI/CD | model evaluation, vector DBs, AI agents |
| Mindset | Writing strict control flow | Model/System design, guardrails, fallback handling |
Basically, it reads as taking on a much more umbrella responsiblity- entry level SWE at my job meant focusing on writing good code, but now I will be expected to handle AI system design and governance.
Key Challenges & Lessons Learned (So Far)
- Convincing Skeptics: You probably have a stereotype in your head that stakeholders want to throw in AI as a buzzword and use it in any way possible. I had the opposite struggle at my company. Even before I was advocating for myself, a previous ex-coworker was trying to push for AI and was always shut down due to security concerns, an old fashioned mindset, and so on. I came in advocating with trepidation in the back of my mind due to this… but the key point that helped me push forward anyways was “answering” the concerns I noticed were being made before they could even be repeated. That truly worked, when they knew their concerns were addressed.
- Applying AI in a useful manner: You should be careful with throwing AI at anything and everything- companies found out the hard way by doing things like replacing customer support teams with AI chat bots. It just doesn’t work to throw an LLM and think it can handle everything! There needs to be a specific, niche goal that can be confidently targeted by the AI system rather than slapping it on as a jack of all trades. My manager and I had to think about this, and we ended up demoing an idea that solves a real problem, that would definitely be more effective than a human dispatcher, and solved the problem of dispatchers retiring + new dispatchers not having as much industry knowledge as the retirees.
- Finding the right tech stack: I’ve briefly mentioned this multiple times, but we really needed tools and frameworks that fit our exact needs. What I ended up finding (ML.net) was EXTREMELY lightweight, simple, and classical- it became a thing before the AI hype even started. We did not need some fancy AI API from a 3rd party, or some LLM to do that. Plus, it’s way more lightweight, free, and local. I guess that’s called good governance.
Looking Ahead
I’m excited to lock into my long term career goal of working directly with AI!
- I always loved agentic AI and models that handle automation, so I’m looking forward to working on any project related to automation.
- I recommend any developer to advocate for themselves (whatever your interest or goal is) to your higher ups instead of simply job hopping. I think if you can advocate correctly and convince people, you can get a lot of value for everyone involved including yourself, and it’s far more valuable career wise to show job growth compared to job hopping.