Which direction fits you?
☐ I love asking 'why did this metric move?'       → analyst / product analyst
☐ I enjoy statistics and modelling               → data scientist
☐ I enjoy building reliable systems              → ML engineer / MLOps
☐ I'm curious about building with LLMs           → applied AI roles
You can start broad and specialise within a year or two of work.

Scan 20 recent data job posts and count mentions
SQL: ____   Python: ____   Statistics: ____
Machine learning: ____   Deployment/MLOps: ____
LLM / generative AI: ____   Cloud: ____
Domain (finance, health, retail...): ____
Use the counts to prioritise what you learn next.

# AI tools can draft code; you must verify it.
# Example: check a drafted function with tests before trusting it.
def median(values):
    s = sorted(values)
    n = len(s)
    if n == 0:
        raise ValueError("empty list")
    mid = n // 2
    return s[mid] if n % 2 else (s[mid - 1] + s[mid]) / 2

assert median([3, 1, 2]) == 2
assert median([4, 1, 3, 2]) == 2.5
print("median checks passed")

Evaluating an LLM feature (e.g. summarising support tickets)
1. Build a test set of real examples with expected outputs
2. Define criteria: accuracy, completeness, tone, safety
3. Score outputs (human review + simple automatic checks)
4. Compare prompt or model versions on the same set
5. Monitor live outputs and collect feedback
Data science skills — sampling, metrics, bias checks — apply directly.

from statsmodels.stats.proportion import proportions_ztest, proportion_confint

# conversions and visitors for control (A) and variant (B)
conversions = [420, 468]
visitors = [10000, 10000]
z, p = proportions_ztest(conversions, visitors)
print(f"z = {z:.2f}, p = {p:.3f}")
for name, c, n in zip(["A", "B"], conversions, visitors):
    lo, hi = proportion_confint(c, n)
    print(f"{name}: {c/n:.2%} (95% CI {lo:.2%} – {hi:.2%})")

Portfolio mix for the next few years
1. A business analysis with clear recommendations (SQL + statistics)
2. A predictive model deployed as an API with monitoring notes
3. An experiment or causal analysis
4. An LLM-based feature with a proper evaluation set
5. One project deep in a chosen domain

Two-year direction plan
Current role/stage: ______
Target direction: analyst / data scientist / ML engineer / applied AI
Target domain: ______
Skills to add this year:
1.
2.
3.
Projects to build:
1.
2.
Review every quarter.

Weekly learning rhythm
2 hours: fundamentals (statistics, SQL, ML theory)
2 hours: building (project work)
1 hour: reading (papers, engineering blogs, domain news)
1 hour: sharing (write-up, talk, or mentoring someone)

Skill radar (0-3)
SQL ___  Python ___  Statistics ___  ML ___
Deployment ___  Cloud ___  LLM evaluation ___
Domain knowledge ___  Communication ___
Focus next on the two lowest that matter for your direction.