CRM workspace
112 synthetic leads, 52 companies, deals, activities, advanced views, and record-level history.
Try it Portfolio project · interactive demo
Revenue operations that can explain itself.
OpsForge is a demo-first CRM and automation platform built to show how I approach data quality, lead routing, workflow reliability, and operational visibility.
My roleArchitecture · Product design · Full-stack implementationContextFictional modern B2B SaaS operations team
99.4% automation success
94.2 data quality
112Synthetic leads
52Companies
11Integration adapters
10Operational SOPs
0Credentials required
The problem
Revenue teams lose trust when data and automation fail quietly.
Leads arrive from disconnected sources with duplicates, missing fields, inconsistent formats, and no reliable owner. Follow-up slows down, attribution becomes questionable, and failed workflows remain invisible.
Messy intakeManual cleanupWrong routingSilent failure
What I built
One operating layer for clean data, fast decisions, and safe recovery.
Each capability is interactive in demo mode and honest about what is simulated.
Data quality
Weighted quality score, review queue, normalization standards, and gated duplicate merges.
Try itLead scoring
A deterministic 0–100 model with every factor and point visible to operators.
Try itRouting engine
Editable, ordered territory and fit rules with capacity-aware assignment architecture.
Try itAutomation control
Trigger-condition-action builder, health metrics, retries, and dead-letter recovery.
Try itWebhook center
Signed endpoints, event logs, latency, response codes, retries, and safe demo replay.
Try itArchitecture
Business rules at the core. Replaceable adapters at the edge.
Lead sourcesForms · ads · referrals · imports
Intake controlsValidate · normalize · deduplicate
OpsForge engineScore · qualify · route · audit
Provider adaptersCRM · email · Slack · calendar
Next.jsReactTypeScriptTailwind CSSZodReact Hook FormRechartsPostgreSQL / Supabase architecturen8n templatesREST & webhooks
How AI is used
Helpful context, not hidden authority.
AI can draft a lead summary, surface likely objections, and recommend a next step. It does not set the deterministic score, merge records, or make the final routing decision.
Structured output with schema validationSeeded fallback when no provider is connectedEvery AI surface labeled as simulated or connectedHuman review for destructive or ambiguous actions
What I learned
Great operations software makes the exception path as intentional as the happy path.
Normalize before matching; otherwise duplicates are a formatting problem disguised as an identity problem.
Scoring and routing earn trust when operators can inspect why a result happened and override it safely.
Retries without a dead-letter queue only delay invisible failures. Recovery needs ownership and context.
Explore the work
See the operating system, not just screenshots.
Open the interactive demo, move a deal, inspect a score, resolve a quality issue, or replay a synthetic webhook.