Case Study · 01Boldtech-class · Retool / internal tools discipline
Internal operations platform for a growth-stage fintech
A high-growth fintech had grown its ops function faster than its tooling. Forty-plus overlapping admin panels, Retool apps, and scripts spread across Ramp-style workflows, Vimeo-scale content moderation, and Bilt-style rewards. Nothing was audited. Nothing was governed. New hires took weeks to become productive, and outages routinely traced back to an unowned tool.
Positioned as: the senior engineer companies bring in when their internal-tools sprawl has become a business risk.
Problem solved
- Fragmented internal tooling with no single owner, no consistent auth, and no audit trail on privileged actions.
- Data access defined by whoever had originally built the panel - not by role or policy.
- Every new operational workflow required a new one-off tool, and nothing decommissioned.
Key features
- Unified internal-ops platform with SSO, RBAC, and mandatory audit logging on every mutation.
- Typed API layer over the operational database with row-level security and per-tenant scoping.
- Component library for ops surfaces - tables, filters, bulk actions, approval flows - built once, reused across teams.
- Governed data-export flow replacing ad-hoc CSV downloads that had bypassed compliance.
- Observability: structured logs, action-level metrics, and alerting on privileged operations.
Technical & strategic contribution
- Owned the architecture end-to-end: authentication, data access, UI patterns, deployment, and runbooks.
- Ran a three-month decommission plan: forty tools consolidated into one, with zero unplanned downtime.
- Trained the internal engineering team on the platform's extension model - new ops workflows now ship in days by product engineers, not weeks by contractors.
Why it matters
This is the work that shows up when a company crosses the point where informal tooling becomes a liability. It requires senior judgment on data access, blast radius, and organizational politics - not just React and TypeScript.
Case Study · 02Datacose-class · AI Employees / operations systemization
AI Employee layer for a service business at scale
A service business north of $10M in annual operations was still routed through its founder. Every important workflow - intake, quoting, scheduling, follow-up, invoicing - passed through one person's inbox. Growth had capped. Selling the business was off the table. The mandate: make the operation run without the owner, using one clean CRM and a governed layer of AI workers.
Positioned as: the engineer founders bring in when the business only runs because they're still in it.
Problem solved
- Founder was the bottleneck for every business-critical decision and every non-standard case.
- Institutional knowledge existed only in Slack threads, email chains, and one person's memory.
- Attempts to hire out of the problem had failed - new headcount added cost without absorbing the load.
Key features
- One system of record: Salesforce reshaped as the operational backbone, replacing five overlapping tools.
- AI intake worker qualifying inbound requests, drafting responses, and creating structured records.
- AI scheduling worker coordinating field teams against real availability, constraints, and customer preferences.
- AI follow-up worker driving quote-to-close and post-service loops, with escalation paths back to humans.
- Governance layer: every AI action logged, reversible, and gated by policy - legal signs off, ops trusts it.
Technical & strategic contribution
- Rewrote the operational model with the founder before writing code - clean processes first, automation second.
- Designed the CRM data model to survive the next five years, not just the current workflow.
- Built the AI worker pipeline as typed, testable, observable services - not ChatGPT wrappers.
- Delivered in the promised 90-day window, with the founder progressively removed from daily operations.
Why it matters
This engagement compresses the whole practice - CRM architecture, AI automation, and business operations - into a single outcome the founder can feel: the business runs whether they're there or not.