Data science isn't one career anymore. It's seven. The 2024 market has fractured into distinct specializations — analytics engineer, ML engineer, MLOps engineer, applied scientist, decision scientist, GenAI engineer, and research scientist — each with different toolchains, interview loops, and salary bands ranging from $115K to $245K median base. If you're still studying 'data science' as a monolith, you're preparing for a job that doesn't exist.
The Seven Roles and What They Actually Do
| Role | Core Focus | Must-Have Stack | Median Base (US) |
|---|---|---|---|
| Analytics Engineer | Modeled data, dashboards, self-serve | dbt, SQL, Looker/Tableau, Airflow | $125K |
| ML Engineer | Productionizing models, scaling inference | Python, PyTorch/TensorFlow, Kubernetes, ONNX | $165K |
| MLOps Engineer | CI/CD for ML, feature stores, monitoring | Terraform, Kubeflow/MLflow, Prometheus, Feast | $155K |
| Applied Scientist | Novel modeling for specific domain problems | PyTorch, JAX, C++, domain expertise (bio, NLP, vision) | $185K |
| Decision Scientist | Causal inference, experimentation, product strategy | R/Python, SQL, stats (causalimpact, DoWhy), Amplitude | $145K |
| GenAI Engineer | LLM app architecture, RAG, eval frameworks | LangChain/LlamaIndex, vector DBs, OpenAI/Anthropic APIs, evals | $175K |
| Research Scientist | Publishing, novel architectures, long-horizon | PyTorch/JAX, distributed training, arXiv track record | $245K |
The Skill Stack That Transfers Everywhere
Regardless of specialization, three competencies separate seniors from juniors in 2024:
Second: version control for data and models. Git is table stakes. DVC or LakeFS for data versioning. MLflow or Weights & Biases for experiment tracking. If you can't reproduce last month's best run, you're not engineering.
Third: communication that drives decisions. The best data scientists I've hired write one-pagers that answer: what did we learn, what should we do, and how confident are we. They present uncertainty honestly — confidence intervals, not point estimates.
💡
TipBuild a portfolio project per target role. Analytics engineer? Deploy a dbt project with tests and docs to Snowflake. GenAI engineer? Ship a RAG app with eval metrics. One polished project beats ten notebooks.
Interview Loops by Role
Analytics engineers get SQL window functions, dbt modeling exercises, and stakeholder simulation calls. ML engineers face system design (feature store, model serving, drift detection) plus coding (custom PyTorch modules, distributed training). MLOps candidates debug broken pipelines and design CI/CD for retraining. Applied scientists present a paper and extend it. Decision scientists run through an A/B test design critique. GenAI engineers build a prompt chain with evals in 90 minutes. Research scientists defend their dissertation and whiteboard a novel architecture.
"The market doesn't reward generalists anymore. It rewards T-shaped people: deep in one specialization, fluent in the adjacent two.
— Hiring manager, FAANG ML platform team
Salary Reality Check
Levels.fyi and Robert Half 2024 data show wide spreads within roles. An ML engineer at a Series B startup: $140K base + equity. Same title at Meta: $190K base + $300K RSUs over four years. Location matters less than company stage and revenue per employee. Negotiate equity refreshers and refresh cycles, not just base.
⚠️
WarningTitle inflation is real. 'Senior Data Scientist' at a 50-person company often maps to 'Data Scientist II' at Big Tech. Level by scope: do you own the problem definition, or just the modeling?
Your 90-Day Plan
Week 1-2: Pick one specialization. Not 'data science.' Read 10 job descriptions for that title at companies you'd join. Extract the repeated requirements.
Week 3-6: Build the portfolio project. Deploy it. Document the architecture decisions. Write the one-pager.
Week 7-10: Mock interviews. Use Pramp, interviewing.io, or peers. Focus on the specific loop for your target role.
Week 11-12: Apply to 20 companies. Track response rates. Iterate resume and project narrative weekly.
✦
The fragmentation is permanent. Specialization is the only path to leverage. Pick your lane, go deep, ship proof, and negotiate like you know your worth — because in 2024, the market finally does.
