US Study vs Work: Strategy | Generated by AI
Question: Should I go to the USA to study or work?
Given your profile (11 years engineering experience, AI infra direction, self-taught background, mortgage/family responsibility, LLM training experience), I would not recommend going to the USA primarily as a student right now. Your higher ROI path is likely:
First get a US-facing AI engineering role / contract / remote opportunity → then use that path for immigration.
1. Studying in the USA: pros and cons
Pros
A US master’s degree can solve:
- No bachelor’s degree problem
- Visa pathway (F-1 → OPT → H-1B)
- Access to AI labs and companies
- Network
Especially schools around:
- AI systems
- distributed systems
- ML infrastructure
- computer science
could help.
Cons for you
The cost is huge.
Typical:
Tuition:
$40k-$80k/year
Living:
$20k-$40k/year
Total:
$60k-$120k/year
A 2-year master’s:
$120k-$240k
≈ 850k-1.7M RMB
You already have:
- mortgage ~900k RMB
- family
- children
Taking another million RMB debt is risky.
Also, you are not a fresh graduate.
Your advantage is:
11 years engineering experience
+
real production systems
+
trained GPT-2 models
+
vLLM/nanoGPT experiments
+
GPU infra experience
A 22-year-old student needs a master’s to prove ability.
You need proof of output, not another classroom.
2. Working in the USA: higher upside
Your strongest asset is unusual:
Many AI applicants:
paper
+
coursework
+
toy projects
You have:
Java backend
distributed systems
cloud
production experience
+
LLM training
+
GPU debugging
+
AI infra experiments
The US market needs:
- ML infrastructure engineers
- inference engineers
- distributed training engineers
- AI platform engineers
Examples:
- vLLM
- Kubernetes GPU scheduling
- CUDA optimization
- training pipelines
- inference serving
Your profile can move toward:
Backend Engineer
|
v
AI Platform Engineer
|
v
ML Infra Engineer
|
v
Research Engineer
3. The biggest problem: degree
Your missing degree is a real constraint.
Some companies:
Google
OpenAI
Anthropic
NVIDIA
Meta
may filter.
But many startups do not.
For AI startups, GitHub matters more.
A strong portfolio:
github.com/lzwjava
projects:
1. nanoGPT training from scratch
2. vLLM optimization experiments
3. distributed training examples
4. CUDA kernels
5. inference benchmark system
6. AI agent infrastructure
can substitute partially.
4. My recommended 24-month strategy
Phase 1: next 6 months
Do not leave without a landing plan.
Build US-market portfolio:
AI Infra Engineer Portfolio
Example:
Project 1
Build mini-vLLM:
tokenizer
↓
prefill
↓
KV cache
↓
continuous batching
↓
scheduler
↓
CUDA kernel
Project 2
Training system:
PyTorch
DDP
FSDP
DeepSpeed
Ray
Kubernetes
Project 3
Production AI service:
FastAPI
vLLM
Redis
Postgres
K8s
GPU monitoring
Phase 2: Apply
Target:
- US startups
- remote-first companies
- Singapore/HK companies with US connections
Roles:
Search:
ML Infrastructure Engineer
AI Platform Engineer
Inference Engineer
LLM Engineer
Research Engineer
Phase 3: immigration
Possible routes:
H-1B
Company sponsors you.
Hard but possible.
O-1 visa
Actually interesting for you.
Requirements:
- open-source contributions
- publications
- talks
- media/blog impact
- judging/reviewer roles
Your blog/community can help.
EB-2 NIW
Also possible if you build:
AI infrastructure contributions
+
papers
+
open source
+
industry impact
5. When would I choose studying?
I would choose USA master’s if:
- you cannot get interviews because of degree filter
- you can enter a strong program
- funding is manageable
- family situation allows
Good targets:
- Georgia Tech OMSCS (online, cheaper)
- UIUC MCS
- UT Austin MSCS
- top AI programs
Avoid:
unknown university
+
$100k debt
+
generic CS degree
That is a bad trade.
My recommendation for your exact situation
I would rank options:
| Path | Score |
|---|---|
| Build AI infra portfolio + US job search | ⭐⭐⭐⭐⭐ |
| Singapore/HK AI infra job first | ⭐⭐⭐⭐ |
| US master’s at top school | ⭐⭐⭐ |
| Random US master’s | ⭐ |
| Stay in legacy Java contractor work | ⭐ |
Your next 12 months should look like:
2026 Aug
|
+-- leave legacy contractor
|
+-- AI infra projects
|
+-- publish technical writing
|
+-- contribute open source
|
+-- apply globally
|
2027
|
+-- AI infra role
|
+-- immigration path
Your strongest move is not becoming a student again. It is becoming a visible AI infrastructure engineer. The USA rewards that profile.
Generated by AI. Curating and sharing still takes effort.
If you find it useful, feel free to
donate.
WeChat: @lzwjava
·
X: @lzwjava ·
Say hi 👋