Field Guides / Breaking into technology / AI / machine learning
Track B · last checked November 19, 2026
AI / machine learning
You like math and experiments, and you are comfortable not knowing whether something will work.
The work: Prepare data, train or adapt models, measure them honestly, and turn them into something a product can use. Most applied jobs are closer to software engineering than to research.
Good fit if you’ve been: Anyone strong in math or statistics, or already working in software or data.
Try it first
Experiment (2–5 hours): Train a simple text classifier with scikit-learn and compare it against a naive baseline. If this is your first one, do the foundations first.
Learn in this order
| Learn | What it unlocks |
|---|---|
| 1. Python, NumPy, pandas and SQL; algebra, statistics and probability | Understand datasets and model behavior |
| 2. Train classifiers with scikit-learn; baselines, train/test splits, precision and recall | Evaluate a model without fooling yourself |
| 3. PyTorch, embeddings, transformers and retrieval | Understand how modern AI systems work |
| 4. Serve a model behind an API with tests, monitoring and cost measurement | Evidence of applied AI engineering |
Portfolio project: A classifier or retrieval system on a public dataset: the data licence, a baseline, a proper train/test split, results with a confusion matrix, an error analysis, limitations, and a small API that serves it. You should be able to explain every part of it, reproduce the results and discuss its limits.
Credentials and degrees
Credentials: Few credentials carry weight in machine learning hiring; projects and math do. Azure AI Fundamentals (now exam AI-901) is AI literacy, not an engineering qualification. AWS Certified Machine Learning Engineer – Associate helps after real AWS practice.
Degree: Computer Science for flexibility; math-heavy CS, Data Science or Statistics if you want research. Research scientist roles usually expect graduate study. How to compare schools.
Where it leads
First jobs to search for: Junior software or data engineer, analytics engineer, ML or AI engineering intern or associate where offered.
Later: ML engineer, AI engineer, applied scientist; research scientist (usually with a graduate degree).
Reality check: Few employers hire people straight into machine learning engineering without software or data experience. The usual route runs through software engineering or data analytics first.
Related DoD work roles (orientation only; about DoD 8140): 623 AI/ML Specialist; 423 Data Scientist; 621 Software Developer.
| Occupation | U.S. 25th percentile | U.S. median | Utah 25th percentile | Utah median |
|---|---|---|---|---|
| Software developers | $105k | $136k | $101k | $129k |
| Data scientists | $86k | $120k | $79k | $108k |
| Computer and information research scientists (often graduate study) | $104k | $140k | $78k | $95k |
BLS May 2025 occupational wages via O*NET, all experience levels, rounded to $1,000. The 25th percentile is not a starting salary; first offers are often lower. No clearance, bonus or equity premium included. How to read these numbers.
Free places to start
Next moves
Enjoyed it? Stay and build the portfolio project. Enjoyed part of it more? Try a neighbor:
- Data & analytics: Finding patterns and explaining the evidence.
- Software engineering: Building applications and fixing bugs.