Every intro, backed by a number.

Assembly Connection reads startup and investor profiles as vectors, scores their fit with cosine similarity, and only introduces the pairs that clear the bar. No cold decks, no spray-and-pray.

0.60
avg. similarity of
generated matches
214
profiles
embedded
47
matches
generated
illustration — not live data

The match happens before anyone says hello.

Three steps run in the background the moment a profile is created. Nothing is introduced until the math agrees.

01

Structure the profile

Thesis, stage, check size, sector focus, or traction and roadmap — captured as structured fields, not a static deck.

stage: seed · sector: climate
02

Embed it

Every profile is converted into a vector via an open-source embedding model — a coordinate in a space where similar intent sits close together.

[0.021, -0.183, 0.077, …]
03

Score and route

Cosine similarity ranks every possible pair. Only scores above threshold get surfaced — to both sides, at the same time.

similarity: 0.91 → intro sent

Built for both sides of the table.

Stop guessing which investor actually wants what you're building.

You fill in one structured profile once. Assembly Connection compares it against every active investor thesis in the system and only routes you to the ones with real alignment on stage, sector, and check size.

  • 01One profile, ranked everywhereUpdate your traction once — every match re-scores automatically.
  • 02See the score before you send anythingKnow your fit before you spend a founder's most limited resource: attention.
  • 03No warm-intro taxMatching runs on thesis alignment, not who you already know.
Northbeam Ventures
Seed · Climate, Industrial
0.93
Alder Capital
Seed–A · Climate
0.88
Fielding Partners
Pre-seed · Deep tech
0.79
top 3 by weighted scoreexample output

Your thesis, applied to every deal in the pipeline at once.

Write your thesis as structured criteria instead of a paragraph on a website. Assembly Connection scores every incoming startup profile against it continuously, so your pipeline stays sorted without a single associate reading a deck.

  • 01Thesis as data, not proseStage, sector, geography and check size become filters that actually filter.
  • 02Ranked deal flow, not raw volumeSee the top-scoring startups first, every time a profile updates.
  • 03Adjustable thresholdWiden or tighten the similarity bar as your fund's stage or focus shifts.
Verdant Robotics
Seed · Raising $3.5M
0.91
Hallow Bio
Seed · Raising $2M
0.86
Current Grid
Pre-seed · Raising $1.2M
0.81
top 3 by weighted scoreexample output

Your next best match is already in the system.

Set up a profile in about ten minutes. We'll tell you your score before we tell anyone your name.