The matching engine

How Continuum works

Every mentor-mentee pair is scored across 40+ signals before it surfaces as a match. This page walks through what the algorithm sees, why each signal exists, and where you stay in control.

Step 1

You build a profile worth matching against

The match engine is only as smart as the data it has. The first 5 minutes after signup are about telling Continuum who you are and what you're after — concretely.

1

Pick a role

Mentor, mentee, or both. Determines which directions the algorithm will surface candidates from.

2

Specialty + career stage

Drawn from a controlled list of 40+ specialties and 8 stages (pre-clinical through executive). Drives the highest-weighted score.

3

Optional identity opt-ins

IMG status, first-generation physician, URM self-identification, languages spoken. Strictly opt-in. You control whether these are used only for matching, displayed on your profile, or kept private (the database enforces this at the view layer — not just the UI).

4

A short bio + topics

Mentors: "Why I mentor" (≤150 char surfaced on match cards). Mentees: a few sentences about where you are. Free text — not parsed as a hard signal, but visible to candidates evaluating you.

5

Optional: NPI verification

One-click check against the federal NPPES registry. Adds a "✓ Verified" badge on your match cards — a strong trust signal for mentees deciding whether to reach out.

Step 2

The algorithm scores every potential pair

For each candidate, Continuum computes a compatibility score from 0–100. Higher = stronger fit. Each signal contributes weighted points; the total runs through a sigmoid normalization so the top of the list is meaningfully separated.

Specialty affinity 0–35 pts
Exact match (35), aspirational student→mentor pairing via interested_specialties (32), same discipline (22), strong cross-discipline affinity like Ortho↔PT (14), mild affinity (8).
Career-stage gap 0–25 pts
Ideal mentor-mentee gap is 2–4 stages. Same-stage peer pairings get partial credit. Too-large gaps are penalized (a department chair isn't usually the right mentor for an MS1).
Geography 0–12 pts
Same city, same metro area, same state, same region. Lighter weight than specialty because remote mentorship works — but in-person matters at decision points (rotations, interviews, networking).
Shared training pathway 0–8 pts
Both navigated the IMG pathway, both first-gen physicians, both did a research year, etc. Only counts when both parties have opted in.
Language overlap 0–4 pts
Speaking the same language matters when discussing nuanced career advice or personal context. Counts up to 3 shared non-English languages.
Mentor capacity filter + boost
Mentors set a soft cap on how many active mentees they can take. Mentors who are at capacity get filtered out of new mentee feeds; mentors with open slots get a small surface boost.
Feedback memory -10 to +10 pts
When you up-vote a match, similar candidates surface higher next time. When you down-vote with a reason ("specialty mismatch", "stage gap off"), that dimension is weighted slightly down for future suggestions to you.
Research / publication overlap 0–8 pts
When CV data is uploaded, the keywords feed a soft research-interest match. Especially valuable for research-focused fellowships and academic-track mentorship.

Plus ~30 more signals — credentials overlap, mentor banner activity, communication preference compatibility (sync vs async), institutional ties, mentee target year vs mentor cohort year, and more. The full list lives in the matching code at site/supabase.js — it's a public JavaScript file. Continuum's algorithm is auditable; we'd rather you understand what we're doing than guess.

Step 3

You see ranked candidates — with the reasoning

The dashboard surfaces your top matches with both the score and the reasons. Each card shows the top-3 reasons the algorithm picked this pair (e.g. "Same specialty", "Ideal experience gap", "Shared training pathway"). The reasoning is part of the product, not a black box.

What it isn't: The score isn't a "compatibility" rating in the personality-quiz sense. It doesn't predict whether you'll get along socially. It predicts whether the *mentorship* will land — i.e., does this mentor have the relevant background, capacity, and stage gap to actually be useful to you right now.

Two highly-compatible people can still not click. The algorithm gets you to the right candidates; the first conversation is where you find out if it works.

Step 4

You reach out — we get out of the way

First message is the activation moment for the whole mentorship. The platform gives you 3 calibrated starter templates if you want them, the messaging UI to send, and structured tools to keep the relationship moving — but the actual mentorship happens between two humans on their own terms.

1

Send a first message

Direct, casual, or specific — pick a starter template or write your own. Most successful mentorships start with a clear 15-minute call request.

2

Formalize when it sticks

Once both of you want a real mentorship, click Formalize. This unlocks session-note logging, pathway visibility (mentor can see your career checklist), and a clear "ended" state when the time comes.

3

Log sessions, capture context

After a call, log a session note: what you discussed, what you committed to, who's doing what before next time. Both parties see all notes; only the author can edit theirs. Recall context months later without rifling through email.

4

Health signals tell you when to re-engage

If 30+ days pass with no message or session, the mentorship card flips to "Cooling." 60+ days and it's "Dormant." Both parties see the same signal — either can reach out to revive.

Step 5

The data improves the platform — your data stays yours

Continuum learns from aggregate platform behavior: which mentor-mentee pairings sustain past a year, which specialties drift toward each other, which dimensions of compatibility predict actual mentorship outcomes. Individual data is never sold or shared with third-party marketers.

Concretely: the matching algorithm gets re-tuned periodically based on which pairs led to active formal mentorships. Aggregate data, de-identified. Your bio, messages, and identity opt-ins are never used for any commercial purpose. See the Privacy Policy for the full inventory.

What we don't do: sell your data, train external AI models on it, target ads at you, share your identity opt-ins with anyone who hasn't been explicitly cleared by your visibility setting.

See it for yourself — free, ~5 minutes

Sign up, fill out the basics, and we'll surface your top matches across mentors, mentees, and opportunities within minutes. No credit card.

Get Started Free → Apply as a mentor