I design and build the AI-driven automation, enrichment, CRM, and analytics infrastructure behind modern go-to-market teams — turning manual sales & marketing motions into production systems that operate unattended, at scale, with duplicate-safety and deliverability built in.
I sit at the intersection of engineering and revenue. I design the pipelines that connect data providers, CRMs, sales-engagement tools, and analytics — then write the code, queries, and orchestration that make them run without hand-holding. From lead sourcing and enrichment to outreach, reply handling, lifecycle automation, and reporting, I own the systems end to end.
My edge is that AI is native to how I build. Custom enrichment agents, LLM-based reply classification, and automated content pipelines are core tools, not add-ons — so one engineer with the right systems can ship what used to take a full RevOps team. I care as much about the unglamorous parts — deduplication, deliverability, rate limits, idempotency, human-vs-bot data integrity — as the flashy ones, because that's what keeps revenue systems trustworthy.
Salesforce-certified. Lifecycle & deal automation, lead scoring, reply-gating, ownership/routing rules, and bidirectional cross-object sync.
Multi-mailbox sequences, compliance-aware enrollment, pacing/warming, and self-healing enrollment pipelines.
Custom enrichment agents, reply classifiers, and multi-stage content pipelines with fingerprint memory loops.
ICP sourcing, contact/company enrichment, verified-email gating, and query-driven reporting (SOQL / SQL).
Mailbox validation, SPF/DKIM/DMARC, inbox-placement safeguards, and bot-vs-human data integrity.
Live GTM dashboards on the edge, webhook orchestration, and integrations across a dozen platforms.
An end-to-end system that enriches a raw company, enrolls it into multi-mailbox outreach, classifies every reply, and routes qualified leads into the CRM — with no manual handoffs.
An agentic pipeline that finds a prospect's employer and a verified email from just a name — combining LLM research agents, people-search APIs, and deterministic pattern-guessing.
An LLM gate that reads every inbound reply, separates real humans from auto-replies and bounces, and only then touches the CRM — with human replies routed to sales in real time.
A button-triggered email broadcast system for product updates and newsletters, sending to segmented CRM audiences with duplicate-safety as the #1 design constraint.
A closed-loop content system: market-intelligence scanning → strategy briefs → AI writers/editors → performance analysis, feeding results back into the next cycle.
Deliverability tooling that catches dead mailboxes standard verifiers miss — the difference between an inbox and a hard bounce.
Leading the migration of the organization's single source of truth from one CRM to another inside a shared enterprise Salesforce org — redesigning lifecycle logic and rebuilding every integration.
Edge-hosted dashboards that give the revenue team a single live view of outbound, broadcasts, content performance, and pipeline health.
Sales needed verified emails for a niche professional audience, but the initial pipeline found a usable, verified address only about 1 in 8 times (~12%). Bad emails meant bounces, wasted sends, and real risk to sender reputation.
Added an LLM research agent plus people-search and profile-lookup tools to pin down the employer, then built deterministic "guess-and-verify" — name-pattern permutations across the real domain and nickname variants, each checked by an email verifier. Writes are gated so only verified, domain-matched emails are saved and anything uncertain is held for review. A customer-base cross-check keeps existing customers out of cold outreach entirely.
Verified-email hit rate climbed to ~75% — roughly 6× — with near-zero bounces and customers automatically protected. In a head-to-head against a commercial enrichment platform, it won on trust and accuracy.
Every outbound reply is auto-classified so only real humans reach sales. During an audit I found genuine human replies were being silently suppressed. Root cause: the sales-engagement API truncated reply text to ~100 characters — so when a recipient's mail server prepended a security banner, the classifier saw only the banner and labeled real replies as automated.
Switched the classifier to read the full reply from the HTML body (which wasn't truncated), removed the text cap, cleared the suppression ledger for affected messages, and reprocessed them idempotently so nothing double-logged.
Recovered the dropped leads into the CRM as qualified leads and closed the whole class of false-suppression bugs — every future reply is now safe from being lost the same way.
A broadcast system has two trust-killing failure modes: emailing someone twice, and engagement metrics polluted by bots that auto-click links. Both had to be designed out from the start.
Built a "tick" architecture that sends at most one email per tick behind content-fingerprint dedup and a per-send re-check — so a contact can't be emailed twice even under concurrency. Layered in sender-warming tiers and pacing, plus forensic bot-vs-human detection (click-velocity + IP/host analysis) to scrub scanner activity from every metric.
Zero duplicate sends across launches, and dashboards that reflect real prospects — bot filtering removed ~80% of raw click noise, so the team could finally trust the numbers.
Open to Go-to-Market Engineering and Revenue Operations roles. If you're building revenue systems that should run on their own, let's talk.