Automation that just runs, so your team stops losing deals to broken plumbing.
I design and ship the integrations, AI agents, and CRM systems that keep marketing and revenue operations honest — quietly, reliably, in production, with the AI stepping in exactly where it earns its place.
What this skill set builds
Patterns pulled from the systems below — the kind of work a new engagement usually falls into one of these.
CRM Data Quality & Attribution
Deduping, multi-source waterfall enrichment, and AI-normalized categorization against your own taxonomy — not generic platform defaults.
Replacing Paid Tools with Owned Infrastructure
Rebuilding a vendor tool's core mechanism natively when the vendor version is unreliable, overpriced, or doesn't fit.
AI-Augmented Sales Enablement
Call and transcript analysis, lead scoring, and coaching insight grounded in real methodology and real business context.
Human-in-the-Loop Ops Tooling
Guided, self-service flows that replace manual multi-step processes prone to human error.
Custom Chat-App Backends
A full interactive Slack (or similar) app backend — events, interactivity, dynamic dropdowns — built without a dedicated server.
Multi-Source Enrichment Pipelines
Chaining several APIs and data sources with automatic fallback when one comes back inconclusive.
Content-Ops Automation
CMS sync and AI-assisted content generation and publishing pipelines.
Lead-Gen Scraping & Qualification
Scraping, AI-based qualification scoring, and handoff straight into Slack or the CRM.
Systems, shipped into production
The mechanism, the stack, and the outcome are exactly what shipped.
Replacing a Paid Sales-Engagement Tool's Core Mechanism
A B2B SaaS company relied on a sales-engagement platform (functionally equivalent to Groove by Clari — market-priced around $50–150/user/month standalone, $200+/user/month bundled) to track sales calendars and log meetings against CRM opportunities. The connection was unreliable: it dropped periodically, missed meetings, and required reps to manually reconnect. Opportunities couldn't be reliably traced back to the meetings that generated them, and BDR compensation — partly tied to tracked meeting volume — became inaccurate whenever the sync silently failed.
What was built
- Detects new and changed calendar events automatically, in real time
- Never logs the same meeting twice, even if the same update arrives more than once
- Filters out notetaker bots and meeting-room resources, so only real attendees get logged
- Automatically finds or creates the right contact record for external meeting guests, checking multiple data sources so nothing falls through the cracks
- Correctly groups related meetings when more than one rep is on the same call
- Keeps its own calendar subscriptions alive automatically, with a daily health check
Meeting-to-opportunity tracking became reliable and gap-free, restoring trust in both BDR compensation data and pipeline attribution.
AI-Normalized Lead Attribution at Scale
The company needed lead-source categorization tailored to its own specific marketing and sales activities. HubSpot's default source-tracking and native automations weren't granular or reliable enough to reflect how the business actually generates and works leads.
What was built
- A single intake point for all HubSpot activity, which automatically sorts each event and routes it to the right specialized process
- New contacts are automatically matched to the correct company record, or a new one is created and linked
- An AI model reviews each contact's activity and assigns it to the company's own custom lead-source categories, not just HubSpot's generic defaults
- Existing lead-source data stays clean and human-readable as it changes over time
- Fills gaps where HubSpot's own automatic enrichment doesn't reliably fire on its own
Lead-source data now reflects the business's actual custom taxonomy instead of generic defaults, feeding cleaner data into attribution reporting and sales routing.
RAG-Powered Inbound Lead Scoring & Proposal Drafting
Inbound leads on a marketplace-platform listing service were reviewed and scored manually. By the time someone got to evaluating a lead and drafting a proposal, momentum was often already lost — competitors could respond faster, or the prospect's attention had moved on.
What was built
- Posting a listing link in Slack triggers an AI-driven review
- One AI model scores the lead's fit, reasoning from the agency's real pricing and past project history
- A second AI model drafts a tailored proposal response, referencing comparable past work
- Both results appear directly in the Slack conversation, ready to act on immediately
Lead evaluation and first response went from a manual, delayed process to immediate in-Slack scoring — letting the team act on promising proposals right away.
Multi-Source Corporate Hierarchy Resolution
Sales territories were assigned partly by account, but multinational companies with regional subsidiaries made ownership hard to track — one account for a global brand's regional entity, a separate account for its parent, with no reliable way to tell they rolled up together. This created territory conflicts and unclear account ownership.
What was built
- Checks an official global business registry first
- Falls back to a public knowledge-graph traversal that maps corporate ownership relationships
- As a last resort, an AI model with live web search reasons through tricky, real-world cases — acquisitions, shell entities, ambiguous headquarters
- Each stage is checked against a set of known-correct answers before being trusted
Validated that automated, multi-source corporate hierarchy resolution is feasible for accurate territory and account mapping. Currently a proof of concept, not yet promoted to production.
AI Sales-Call Coaching Analysis, at Scale
Sales leadership needed to understand why deals were stalling versus progressing, at a level deeper than gut feel from spot-checking individual calls.
What was built
- Automatically finds and collects the call recordings tied to each opportunity, only pulling new calls on repeat runs
- An AI model evaluates each conversation, citing specific moments and quotes rather than generic feedback
- Findings are compiled into a clean, readable report for each deal
- Results can also be rolled up across many deals at once into a single comparative summary of what separates progressing deals from stalled ones
Surfaced a clear, evidence-backed pattern: reps were frequently skipping discovery and jumping straight into a features pitch. Run against a real batch of 281 opportunities, with findings directly praised by the CEO, sales managers, and CTO.
Self-Service Demand-Gen Data Import
Getting a new list of leads — from an event, a webinar, anywhere — properly into the CRM required a multi-step manual process: mapping CSV columns correctly, resolving picklist values, creating a matching Salesforce Campaign, and more. People frequently forgot a step or got it wrong, especially since the process wasn't run often enough to become second nature.
What was built
- Submitting a lead list kicks off a step-by-step walkthrough inside Slack — no manual training or CRM knowledge required
- One AI model automatically maps the spreadsheet's columns to the right CRM fields (personal data is stripped out before anything is sent to the AI model)
- A second AI model cleans up messy or inconsistent values automatically
- After a final confirmation, everything is imported and the required CRM records are created automatically — no steps to remember, nothing to get wrong
Replaced a manual, error-prone, easy-to-forget-a-step process with a guided flow that anyone at the company can run correctly.
Have a process that's held together by hope and a spreadsheet?
Send a note with what's breaking and what it's connected to — HubSpot, Salesforce, Slack, or something stranger. If it's a fit, next step is a short call.