Hi HN, we built StartupVCmatch to reduce a painful failure mode in fundraising: spending weeks talking to investors who were never a fit.
What it does
You upload a PDF pitch deck. We parse slide text, extract key fundraising attributes (e.g., sector, stage, geography, check size + other signals), and match them against a VC dataset to produce a ranked list. For each VC we include a reason/explanation for the match so it’s not a black box.
Why we built it
A lot of founders either:
overfit to brand names (“top funds”), or
drown in directories and filters,
and end up with a list that’s not aligned with what their deck actually signals.
Our bet: the deck is already the best “structured description” of the company, so we use it as the input.
What we’d love HN feedback on
If you were a founder, what would make you trust (or distrust) a ranked investor list?
Scoring: should we keep it rule-based + deterministic (easier to explain), or allow more ML/LLM heuristics (potentially higher recall)?
Output format: would you want just the top matches, or “tiers” (must-talk / good fit / exploration)?
Happy to share more details about the extraction schema + matching logic if anyone’s interested.
Hi HN, we built StartupVCmatch to reduce a painful failure mode in fundraising: spending weeks talking to investors who were never a fit.
What it does You upload a PDF pitch deck. We parse slide text, extract key fundraising attributes (e.g., sector, stage, geography, check size + other signals), and match them against a VC dataset to produce a ranked list. For each VC we include a reason/explanation for the match so it’s not a black box.
Why we built it A lot of founders either:
overfit to brand names (“top funds”), or
drown in directories and filters, and end up with a list that’s not aligned with what their deck actually signals.
Our bet: the deck is already the best “structured description” of the company, so we use it as the input.
What we’d love HN feedback on
If you were a founder, what would make you trust (or distrust) a ranked investor list?
Scoring: should we keep it rule-based + deterministic (easier to explain), or allow more ML/LLM heuristics (potentially higher recall)?
Output format: would you want just the top matches, or “tiers” (must-talk / good fit / exploration)?
Happy to share more details about the extraction schema + matching logic if anyone’s interested.