How matching works

We didn't just build another AI matching tool. We built one that understands how healthcare training actually works.

Most "AI mentorship" platforms pair people based on job titles and generic interest tags. That's not mentorship — that's a directory. Continuum is different, because clinicians are different. Every stage of training has its own rhythms, its own gatekeepers, its own anxieties. Our matching algorithm was built to respect that.

The Principle That Shapes Everything

Recent experience matters more than pure seniority.

A current resident who just matched knows more about today's application cycle — the signals, the tokens, the post-Step-1-pass/fail landscape — than a senior attending who matched a decade ago. A fourth-year medical student who just chose a specialty knows more about weighing that decision in 2026 than a mid-career physician who made the same choice in 2012. Our algorithm encodes this principle: it prioritizes the mentor who has just walked the path you're about to walk, while still surfacing deeper expertise when what you need is long-term career vision, not process recency. We believe this is the single most important thing missing from mentorship platforms today.

What the algorithm considers

A multi-dimensional fit score is computed for every potential match. Among the factors weighed:

🎯

Specialty alignment

Direct match, sub-specialty, discipline family, and cross-discipline clinical collaborations — calibrated per specialty so physical therapists aren't only matched with other PTs.

⏳

Career-stage trajectory

The right experience gap. Too close and mentorship loses value. Too far and recency erodes. Our model rewards the specific gap shapes that predict real clinical mentorship outcomes.

🧩

Strength-to-gap complementarity

We analyze each CV to identify where a clinician is strong and where a mentorship could meaningfully move them. Mentors are matched to mentees whose development areas they can actually address.

🔬

Research compatibility

Depth, subject alignment, publication cadence, and active-opportunity signals. Matters especially for trainees considering academic paths or seeking first research roles.

🏥

Institutional and geographic fit

Same health system, same city, same region — useful for in-person mentorship and local networks. Non-penalizing for remote mentorships when that's the preferred format.

📚

Teaching and leadership depth

Separate signals from research. A mentor who is an outstanding educator or demonstrated leader brings specific, measurable value that the algorithm weights accordingly.

🤝

Representation preferences (opt-in)

When you actively choose to be matched with mentors who share elements of your lived experience, we honor that. When you don't, demographic data never touches your matches — full stop.

⚖️

Mentor capacity and availability

A mentor already working with eight mentees isn't the right match for a ninth. The algorithm detects load and protects mentor bandwidth — so every new match gets real attention.

🔁

Continuous learning

Thumbs-up, dismissals, and real conversation signals refine your ranking over time, with a 6-week half-life so stale preferences fade. The algorithm gets sharper the more you use it.

How we validated it

Designed with clinicians

Scoring principles were codified by a practicing clinician, not a data scientist bolting on healthcare tags. Every signal has a clinical rationale.

Tested at scale

Stress-tested on 1,000,000 synthetic healthcare professional profiles across every career stage and specialty — 11 stages × 30+ specialties.

Measured for fairness

Approximately 120 million pair comparisons run through validation. Every synthetic user found viable matches. No edge-case exclusions.

Tuned for recency

Explicit post-simulation tuning to reward near-peer mentorship and surface current-cycle experience over decades-old credentials.

1,000,000

Synthetic profiles tested

~120M

Pair comparisons

Weighted signals

330+

Stage × specialty combinations

What this doesn't claim

Healthcare professionals are rightfully skeptical of AI claims. Here's what we're not saying:

We are not clinically validated. No peer-reviewed study yet — though we plan to publish outcomes as the platform matures.
We don't replace institutional mentorship. Your program director, chief, and attendings know you; we're a complement, not a substitute.
We don't give medical advice. This is a mentorship platform, not a clinical tool. Our community guidelines explicitly prohibit patient-specific clinical guidance between users.
We don't use your CV to train general AI models. Your data improves your matches — and only yours.

Baker Mills, MD, MS

Founder · Orthopedic Surgery Resident, Duke University

"I built Continuum because the gap between the clinicians who need mentorship and the clinicians who could offer it is absurd — and the tools to bridge that gap have been consumer-grade for a decade while the problem itself is anything but. Every design choice in this algorithm is answerable to the question: would a busy resident on call actually benefit from this? If the answer is no, we cut it."

See who your algorithm-matched mentor would be.

Join the waitlist. Early members help shape the product and get founding-member pricing.

Join the waitlist →