TalengineerQuality Score
The score is not a black box
TalScore rolls four verifiable signals into one 0–100 number that employers can sort and gate by. Here is exactly how every point is earned — and how we stop it from being gamed.
Four signals, one number
Each dimension answers a question an employer actually cares about. The weights add up to 100.
Can they actually do the work — not just talk about it?
Every engineer passes a practical AI technical interview before they can be matched. That 0–100 score maps directly onto 25 points. It is the capability baseline; nobody skips it.
Can they prove it under exam conditions?
Points come from timed, proctored certification exams — L1 = 8, L2 = 16, L3 = 25 per track, capped at 25. Self-reported résumé skills earn nothing here. Only what was verified counts.
Did the engineer deliver on time, every time?
The largest slice — real ratings from employers after real jobs. We use a Bayesian average, not a raw one, so a single glowing review cannot game the number.
Do they finish what they start, without disputes?
One point per completed order (capped at 10) plus a 10-point no-dispute bonus. But there is a red line: if more than 1 in 10 jobs ends in a dispute, this entire 20-point dimension drops to zero.
How the points are calculated
| Dimension | Max | How it is earned |
|---|---|---|
| AI Technical Screening | 25 | Onboarding AI interview score (0–100) mapped linearly to 0–25. |
| Platform Certification | 25 | Highest level per track: L1 = 8, L2 = 16, L3 = 25. Summed across tracks, capped at 25. |
| Employer Rating | 30 | Bayesian-averaged star rating (prior: 5 reviews at 3.5★) mapped from 0–5 stars onto 0–30. |
| Reliability | 20 | 1 point per completed order (cap 10) + 10 for a clean record. Dispute rate over 10% → the whole dimension is 0. |
Why a Bayesian average, not a plain one
A raw average lets one 5-star review rocket a brand-new engineer to the top of the list. Instead, every engineer starts as if they already had 5 reviews at 3.5 stars — the platform-wide prior. Real reviews then pull the number toward the truth as they accumulate. A veteran with 40 honest reviews reflects their real record; a newcomer with one review does not get to fake a perfect one.
The dispute red line
Reliability is not a nice-to-have — it is a gate. Completed orders and a clean dispute record build it up, but if more than 10% of an engineer’s jobs end in a dispute, the whole 20-point dimension collapses to zero. No partial credit. On industrial sites, a pattern of disputes is disqualifying, and the score treats it that way.
Four badge tiers
The final 0–100 score maps to a badge employers see at a glance.
A live number, not a monthly snapshot
TalScore is recomputed on the events that actually change an engineer’s standing: every completed order, every new review, and every certification issued. There is no monthly batch job to wait for — the moment the underlying facts move, so does the score.
See the score in action
Employers sort and gate their shortlist by TalScore. Engineers raise it by getting certified and delivering clean, on-time work.