Pacome
A stock-suggestion platform built on machine learning signals — with the data-provider plumbing that actually decides whether fintech products survive.
What they needed — in their own words
“Our models pick stocks. Subscribers should see ranked suggestions, get alerts when a signal changes, and pay monthly. The hard part is always the market data — feeds break, symbols change, providers throttle.”
The system they run on today
- Subscriber dashboard with ranked, explainable stock suggestions
- Signal-change alerting via email and push
- Subscription billing with plan upgrades and dunning
- Market-data ingestion resilient to provider quirks
Arrived, assessed, hardened
Every Neolen engagement follows the same path: an honest assessment of what's actually there, then the hardening work that makes it a system. Here is how it played out on this one.
What arrived
A signal engine that worked and an ingestion layer that trusted the market-data provider completely: no rate-limit handling, no retry policy, and symbols assumed to be stable forever.
What the assessment found
Testing against the provider's real sandbox surfaced the failure the founders had felt but couldn't name: after corporate actions, a ticker rename silently orphaned live signals. Subscribers would simply stop receiving updates for a stock they were paying to follow — with no error anywhere.
What we hardened and shipped
We built a corporate-actions reconciliation layer with symbol lineage tracking, so a renamed or merged ticker carries its signal history forward, and wrapped ingestion in rate-limit-aware retries with alerting when the feed misbehaves.
What the client runs on today
Pacome's subscribers get uninterrupted signals through ticker renames, splits, and provider hiccups — failure modes that would have quietly bled paying customers without anyone knowing why.
Added to the playbook
Symbol-lineage checks joined the audit harness for every application that ingests third-party financial data.
Hit a wall with something you've built?
The audit is fixed-fee and standalone — a plain-language report of what's actually wrong, ranked by risk, yours whether or not you continue.