Engineering
Intelligent
Systems
dancodea.tech
Full-cycle pipeline automation with intelligent email classification
Navigating modern job markets involves tracking dozens of simultaneous opportunities across complex interview stages. Candidates lose track of follow-up deadlines, accidentally submit duplicate applications to the same department, and miss critical recruiter emails buried under generic auto-replies.
We architected an autonomous, multi-tiered CRM engine built on n8n and powered by Groq LLMs and Airtable. The system unifies application intake with fuzzy title matching, executes scheduled SLA-driven follow-up alerts to Discord, and automatically intercepts, parses, and classifies incoming recruiter emails into structured lifecycle events.
Custom JavaScript strips legal suffixes (LLC, Corp, Ltd) and matches company aliases. A Levenshtein distance algorithm fuzzy-compares job titles (85% similarity threshold across a 60-day window) to block duplicate submissions and protect data integrity.
Interactive webhook triggered directly from Airtable buttons smoothly transitions saved opportunities from Discovered to Applied with strict status validation to prevent accidental overwrites on active interview stages.
A scheduled daily monitor audits active applications against stage-specific SLA thresholds (7 days for Applied, 3 days for Interviews). It calculates elapsed time, prevents re-nag spam, and dispatches rich action-item cards directly to Discord.
Monitors Gmail in real-time, pre-filtering recruiting keywords and ATS domains (Greenhouse, Lever, Ashby, Workday). Resolves sending companies, links active thread IDs for zero-latency future routing, and matches target records.
Groq LLMs parse recruiter emails into strict structured schemas (Interview Invitation, Rejection, Info Request, Confirmation). Classifications with ≥75% confidence automatically update Airtable stages (e.g. promoting to Technical Interview on coding signals), while low-confidence events ping Discord for human review.
System Architecture & Workflow Modules
Three-Pillar Unified Automation Engine
Unified orchestration handling intake deduplication (Workflow A), scheduled follow-up cadence (Workflow B), and real-time AI email classification with Airtable state syncing (Workflow C).
Module A: Intake & Deduplication
Form captures opportunity, normalizes legal company strings, and fuzzy-compares position titles to block redundant applications.
Module B: Follow-Up Monitoring
Daily cron engine checks status change dates against SLA windows and fires automated Discord reminders with context.
Module C: Recruiter Email AI
High-throughput Groq LLM analyzes email payloads, gates decisions on 75% confidence, and auto-syncs Airtable interview stages.
Workflow Testing & Live QA Verification
Production testing simulations were performed to validate multi-channel intake deduplication, fuzzy company normalization, scheduled follow-up triggers, inbound Gmail AI classification, and state promotion in Airtable. Below are the verified live execution audit logs and central application records.
1. Airtable Real-Time Event Log & Audit Trail
Live production audit trail tracking execution stages including application_promoted, follow_up_created, duplicate_blocked, and email_unmatched with source channels and exact timestamps.
2. Airtable Central Applications Database
Centralized candidate and opportunity CRM capturing normalized company names, applied job titles, job URLs, submission timestamps, attached resumes, and verified recruiter contacts.
We build bespoke end-to-end operational systems.