Most demanded skills
Share of postings in the latest four quarters that mention the skill. Click one for its trend.
Starting the Python engine in your browser.
Loading Python (about 12 MB, first visit only; later visits work offline)…
Are skill programs teaching what jobs ask for?
SkillAlign reads job postings, works out which skills employers want and which are changing beyond chance, then scores each training program against that demand.
Emerging
Rising by 30% or more, beyond chance, and within the same job roles.
Declining
Falling by 20% or more, beyond chance, and within the same job roles.
Trend
Training programs, scored against the jobs they train for
Each program is compared only with postings for its own target roles. The score weights each in-demand skill by how often jobs ask for it and by whether the program gives it enough hours. The range shows how much the score could move with a different sample of postings.
Gap matrix
Rows are programs; columns are the skills each program's own target jobs ask for most. The darker a cell, the more that program's jobs ask for the skill. A tick means it is taught; a red outline marks a skill in 15% or more of its jobs that the program does not teach.
rarely askedasked in most postings✓ taughtgap: in demand, not taught
Check a syllabus
Paste any course outline, one module per line, with hours if you have them. SkillAlign finds the skills it teaches, tells you which lines it could not read, and scores it against the jobs you name.
How ready am I for a job?
Pick the job you want and tell SkillAlign what you can do. It shows what employers ask for, what you're missing, which words it couldn't read, and which program closes the most gaps.
Use your own job postings
The built-in postings are synthetic. Upload real postings and every score on the site is recomputed from them. Nothing leaves your browser.
CSV format
Required columns: title, date, and a description and/or a skills column (like "Python|SQL|Excel", common in job-board exports). Optional: sector, district (or city), state. Common export names such as posted_date and key_skills are understood. Commas, semicolons, tabs or pipes as separators. Dates as 2025-08-14 or 14/08/2025. Skills are read from the description, never the title. At least eight quarters give trends; fewer still give shares and scores.
Try the skill reader
Type any text to see which skills SkillAlign recognises, the exact words it matched, and the items it could not read. Marathi skill words work too.
How it works
Every number on this site is computed by a small Python program that runs in your browser. There is no AI model and no server: the same data always gives the same answer, and every rule below can be checked by hand.
1. Reading skills out of text
A taxonomy lists 79 skills with 393 ways of writing them, including ITI trades and Marathi words. The text is split into words, and at each word the longest known phrase (up to four words) is looked up in a dictionary. Short ambiguous words only count in their proper form ("Excel", not "excel at"), plurals and hyphens are tolerated, and a skill right after "no", "not" or "without" is skipped. Skills come from the posting description only, never the job title.
2. Measuring demand, honestly
A skill's share is the fraction of postings that mention it, shown with a 95% range. Change compares the latest four calendar quarters with the four before. A change is only called emerging or declining when a two-proportion z-test says it is beyond chance, corrected for testing dozens of skills at once (Bonferroni), and when the same change shows up within job roles, so a shift in which roles are hiring is not mistaken for a skill trend. A straight-line forecast is drawn only with six or more quarters.
3. Scoring a program
Only postings for the program's own target roles count. Skills in at least 15% of them are core. Each core skill earns credit for being taught, scaled by hours: full credit needs at least half the hours its share of demand would justify, so a one-hour "Introduction to X" earns almost nothing. The alignment score is the demand-weighted average credit. Its range comes from resampling the postings 150 times. A score whose range crosses 50 or 75 is marked borderline.
4. Recommendations
- Add: a core skill the program does not teach.
- Expand: taught, but earning under 75% credit because it gets too few hours.
- Prepare: not core yet, but rising beyond chance and forecast to reach 15% within a year.
- Reduce: taught, yet confidently under 5% of postings. Never applied to mandatory modules.
- Review: still asked for, but falling beyond chance within the same roles. Never applied to mandatory modules.
Checked against fake trends
On sample data generated with no trends at all, across 20 random samples, this method reports no emerging or declining skills; the first version reported about one per sample. On trended samples, every trend it reported (42 of 42) was really in the data. It is deliberately cautious, so small changes go unreported. Scores move by a few points between samples, which is what the ranges show.
Limits
Postings say what employers advertise, mostly in the formal sector, not everything a job needs. A skill outside the taxonomy is invisible, so the page lists what it could not read. The built-in data is synthetic; upload real postings for real conclusions.