Tech Demand

What is worth learning · reported week 2026-W37 (last completed ISO week, SGT) · 3-5 yrs · AI-ML
Experience: all0-23-56+unstated
Role: allAI-MLBackendDataFrontendFullstackMobileOther-ITPlatformSRESecurity

1. Demand ranking

27 enriched SWE postings in 2026-W37. Share = postings mentioning the technology ÷ that number — postings still awaiting enrichment are excluded from the denominator, so a processing backlog cannot depress every share at once. The chart shows the top 15; the table in section 3 lists the top 30.

python21machine-learning18llm17docker15rag12azure10generative-ai10aws9fastapi9google-cloud9kubernetes9javascript8sql8deep-learning7react7

2. Momentum (vs the previous 4 weeks)

Heating up

TechnologyShareChangePostings
generative-ai37.0%+14.0pp10
docker55.6%+8.5pp15
azure37.0%+3.8pp10
machine-learning66.7%+3.6pp18
rag44.4%+2.3pp12
llm63.0%+1.1pp17

Cooling down

TechnologyShareChangePostings
python77.8%-6.5pp21

Change is in percentage points of share, not relative percent: a technology going from 1 to 3 postings would otherwise read as +200% and top the board. Boards consider every technology above the bar, not only the 30 the table below shows.

3. Salary premium and entry-friendliness

Premium compares the median advertised monthly salary of postings mentioning a technology against the overall median, over the trailing 90 days. Baseline: S$8000, the median of 322 postings advertising a monthly range — medians pin the unit so the figures are comparable. Separately, 100.0% of SWE postings state pay at all (322 of 322, in any unit); the rest hide it, and no figure here describes them. Entry-friendly is computed over the same 90-day window. Premium mixes seniority in (senior roles name more infrastructure); pick an experience band above to compare within one. Entry-friendly = the share of postings mentioning the technology that ask for at most 2 years' experience, or are Intern/Junior roles with no stated requirement. The table lists the top 30 technologies by postings.

TechnologyKindPostingsShareSalary premiumEntry-friendly
pythonlanguage2177.8% +0.0% 0.0%
machine-learningai1866.7% +0.0% 0.0%
llmai1763.0% +0.0% 0.0%
dockertool1555.6% -9.4% 0.0%
ragai1244.4% +0.0% 0.0%
azurecloud1037.0% -6.2% 0.0%
generative-aiai1037.0% +0.0% 0.0%
awscloud933.3% +0.0% 0.0%
fastapiframework933.3% -6.2% 0.0%
google-cloudcloud933.3% -7.5% 0.0%
kubernetestool933.3% -6.2% 0.0%
javascriptlanguage829.6% -9.4% 0.0%
sqllanguage829.6% -9.4% 0.0%
deep-learningai725.9% -9.4% 0.0%
reactframework725.9% -9.4% 0.0%
nodejslanguage622.2% -9.4% 0.0%
langchainai518.5% -6.2% 0.0%
openaiai518.5% -6.2% 0.0%
pytorchai518.5% -9.4% 0.0%
javalanguage414.8% +12.5% 0.0%
csharplanguage311.1% -9.4% 0.0%
postgresqldatabase311.1% —(n=17) 0.0%
tensorflowai311.1% -9.4% 0.0%
typescriptlanguage311.1% +18.8% 0.0%
airflowtool27.4% —(n=8) 0.0%
angularframework27.4% —(n=3) 0.0%
computer-visionai27.4% +9.4% 0.0%
flaskframework27.4% —(n=14) 0.0%
gittool27.4% -6.2% 0.0%
github-actionstool27.4% —(n=4) 0.0%

4. What else they ask for

These are MyCareersFuture's own skill tags — the competencies the employer filled in on the form, over the trailing 90 days (2026-06-23 → 2026-09-20) across 322 postings. They are not the technology ranking above: languages and frameworks appear only in the free-text description, which is why this system reads it separately. "Must-have" is the share of postings listing the tag that marked it essential rather than desirable — a tag that is everywhere but rarely essential is table stakes, one that is usually essential is a filter someone is applying.

SkillPostingsShareMarked must-have
Python11736.3%7.7%
Computer Science9128.3%3.3%
Machine Learning7021.7%12.9%
AI Agents6821.1%7.4%
PyTorch6520.2%3.1%
Artificial Intelligence6219.3%24.2%
TensorFlow5617.4%1.8%
Design5517.1%0.0%
Ai4614.3%6.5%
LLMs4112.7%9.8%
Prompt Engineering3510.9%0.0%
C++3410.6%5.9%
Data Science3310.2%3.0%
Collaborate With Engineers329.9%0.0%
Programming329.9%0.0%
Numbers computed by SQL from public MyCareersFuture data; data is refreshed daily, so it lags the live market by up to 24h. Methodology: docs/03-data-model.md · data freshness · Compliance: aggregate statistics only, no personal data.