Reply Classifier Cascade
A cost-ordered classifier for English and Dutch cold-email replies that decides which replies become CRM deals, built so a false “hot deal” needs several layers to agree.
- 1The problem
Keyword filters match the quoted thread, miss negation and miss implicit interest; a single model is confidently wrong. A false deal wastes sales time and corrupts pipeline numbers.
- 2Our contribution
Designed and built end to end by Kumar Sahil.
- 3What we built
A five-layer cascade: platform flag (vetoable), anchored EN/NL regex rules, an LLM with a confidence floor, a Naive Bayes fallback that can’t create a deal without a non-negated positive cue, and a keyword safety net that routes to a human. Free deterministic layers settle most replies first, so the paid model only sees the hard ones.
- 4The result
Released as open source. Read the code, tests and design notes on GitHub.
Other systems we’ve built
LLM Gateway
A provider-neutral LLM gateway: code asks for a job, the gateway routes it through a chain of models with safe failover, a deadline per call, and the cost of every call written to a ledger.Read the case studyLLM Eval Gauntlet
Invariant-checked evaluation suites for tool-calling agents that run in CI and fail the build when an agent invents a number, leaks its prompt or obeys an injection.Read the case studyFind the workflow worth automating first.
Tell us about one process that costs your team hours every week. We will tell you honestly whether AI or automation is the right fix, and what a first build would take.