Ask HN: Stuck in Enterprise CRUD: Seeking Path to Hands-On AI Role
2 comments
My suggestion: upskill to match a growing need.
No formal education really exists in this field yet. So don’t get scammed. Instead, look at everything Andrew Ng, Andrew Karpathy, and HuggingFace are doing. The recently published HF book on training from scratch looks interesting, go learn to train some models from scratch. Go put some cash down on OpenRouter and the A/B blind test websites, and play with open models enough to have an overly excited conversation in a bar, waving your beer and pontificating about your favorites and why China is or isn’t winning — in your opinion.
You say you’ve tried side projects and such, but their exact nature will either color you as a normie, or someone who could help a lab make forward progress. Bad: “I typed a prompt and look what I got!” Better: “I used a vibecoding tool to get started then iterated with Claude Code / Codex, pulled in some stuff I did in Figma Make, and got ChatGPT to walk me through setting up the infra to get the app deployed”. Even better: “I wanted to understand nanochat so had Gemini help me walk through it, can I explain it on a whiteboard?” Even better: “I ported nanochat to Rust with the help of Gemini, that was really cool but man the borrow checker gave me grief with X and Y and Z.”
There are people making amazing things at every level of the stack which demands net new hiring (as opposed to hiring to backfill, which is going to zero).
Find where you have the most fun, play there, and match it to some net new hiring need in the industry.
Welcome to an incredible journey!
No formal education really exists in this field yet. So don’t get scammed. Instead, look at everything Andrew Ng, Andrew Karpathy, and HuggingFace are doing. The recently published HF book on training from scratch looks interesting, go learn to train some models from scratch. Go put some cash down on OpenRouter and the A/B blind test websites, and play with open models enough to have an overly excited conversation in a bar, waving your beer and pontificating about your favorites and why China is or isn’t winning — in your opinion.
You say you’ve tried side projects and such, but their exact nature will either color you as a normie, or someone who could help a lab make forward progress. Bad: “I typed a prompt and look what I got!” Better: “I used a vibecoding tool to get started then iterated with Claude Code / Codex, pulled in some stuff I did in Figma Make, and got ChatGPT to walk me through setting up the infra to get the app deployed”. Even better: “I wanted to understand nanochat so had Gemini help me walk through it, can I explain it on a whiteboard?” Even better: “I ported nanochat to Rust with the help of Gemini, that was really cool but man the borrow checker gave me grief with X and Y and Z.”
There are people making amazing things at every level of the stack which demands net new hiring (as opposed to hiring to backfill, which is going to zero).
Find where you have the most fun, play there, and match it to some net new hiring need in the industry.
Welcome to an incredible journey!
Thank you for the detail reply. I feel less overwhelmed now and sort of idea for next step to take. I've noticed Andrew NG deep learning specialization to be frequently mentioned along with Karpthy Zero To Hero free course.
I have toyed around free open source models, gemma3n, and running locally, etc.
As per project I did, it was to load up vector database and built a 'mini' agentic chat bot. Nothing fancy like fine tuning a model.
Ambitious project I've been thinking is to build an inference on top of established model, for example inference on deep seek model or llama in Python or Rust. But I run into self doubt since I don't have any formal ML background, so I'm not sure how I can pull it off.
I do keep an eye on trending github project, I'm amazed on project people make. My recent interest has been coding agent that operates within CLI and coding agent that work off git workspace.
This brings back a point I should have mentioned, since I don't have a picture, I am unable to prioritize on which step to take first and from there, branch off.
I have toyed around free open source models, gemma3n, and running locally, etc.
As per project I did, it was to load up vector database and built a 'mini' agentic chat bot. Nothing fancy like fine tuning a model.
Ambitious project I've been thinking is to build an inference on top of established model, for example inference on deep seek model or llama in Python or Rust. But I run into self doubt since I don't have any formal ML background, so I'm not sure how I can pull it off.
I do keep an eye on trending github project, I'm amazed on project people make. My recent interest has been coding agent that operates within CLI and coding agent that work off git workspace.
This brings back a point I should have mentioned, since I don't have a picture, I am unable to prioritize on which step to take first and from there, branch off.
I'm a developer with six years of experience (backend/frontend/infra), primarily doing vanilla CRUD work in an enterprise setting. I'm struggling to pivot into a more in-demand field like AI.
I've tried upskilling (certs, technical books, personal projects), but without professional exposure to an AI project, my efforts haven't opened doors. I feel stuck.
I see three main paths forward to bridge this experience gap:
1): Apply directly to AI-focused roles, hoping a company takes a chance on a "vanilla Java dev."
2) Go back to school for a Master's degree in AI.
3): Double down on more advanced AI certifications and personal projects.
Fellow Devs: How did you successfully pivot from traditional development into a high-demand field? What was the critical action that finally broke the barrier for you?
+++++ RAW VERSION ++++++++
Hi HN,
I'm doing bit of retrospection from where to go from where I am in my career. I'm struggling a bit to break into industry that is in demand verses what I have been doing - which is CRUD style work in enterprise company. I have six years experience in diverse area, backend, front end, infrastructure. .
I have done my best to upskill which ever way I can, from reading technical books, "AI Engineer" to obtaining certification that is AI focused to better understand AI terminology that is.
What triggered me to make this post is I came across an individual on LinkedIn who has been in product and managed to get into 'AI Project' and now appears to be trying to transition into 'hands on' AI role by going back to school and doing master's degree in AI space.
Compared to my experience, I have been working as a developer for six years. However, I have never been exposed to any real AI project I can work on.
So, for me to take the next step, I can see only three options:
1) applied to AI focus role and 'hope and pray' some AI company is willing to take a risk to get vanilla Java backend dev to build out their AI product
2) Go back to school and do master's degree in AI like the individual did on LN
3) do AI focus certification, like AI Engineer Azure, Coursera/deeplearning, certification and, again, 'hope and pray' it will make my resume stand out?
I'd love to hear from fellow devs who were able to change their trajectory to transition from vanilla development into field that is in demand and open more opportunities?
I feel extremely sideline the way I am in my current position. I genuinely believed doing AI certification and a personal project in AI will open doors, but turns out, it didn't. Also, my company which is financial services space is not open to having AI project deployed. For example, we were very late to adopt copilot. So, that's another limitation.