Roman Licursi

GTM Systems Analyst Intern at Together AI · Computer Science at UW-Madison

I build production AI agents and revenue systems for GTM teams, from Salesforce and Clay automation to evals, entity resolution, data pipelines, and security controls. I focus on systems that are measurable, grounded, and useful inside the rep’s workflow.

26x faster inference 6 live data lanes +7.5pp answer accuracy

Email → LinkedIn → GitHub → Writing → Resume →

AI agentsRevOpsGTM systemsPythonSQLClaySalesforceSupabasedbtSnowflake

Projects

Clay GTM Account Prioritization Workflow May 2026

Featured View public Clay template →

Clay Claygent Formulas Waterfall Enrichment Linked Tables Account Scoring

Find

›

Enrich

›

Transform

›

Export

Built a Clay workbook that finds high-fit software companies, enriches them with hiring, funding, product, and infrastructure signals, then turns the research into a ranked account list for a technical sales team. Claygent handles web research on model usage and infrastructure needs, formulas drive scoring and tiering, linked Jobs and People tables surface buying signals and decision-makers, and a work email waterfall makes the best accounts actionable.

Companies

Fit signals, growth, funding stage, readiness score, tier, why now

Jobs

Recent ML / inference / applied AI hiring as buying signal

People

Decision-makers, email waterfall, opening lines

FETE framework Claygent research Scoring formulas 3 linked tables Email waterfall Outbound-ready shortlist

Salesforce Lead Routing & Scoring Pipeline Mar 2026

View on GitHub →

Python Salesforce simple_salesforce scikit-learn SFDC Flows

End-to-end Salesforce RevOps automation: Python logistic regression pipeline scores 100+ leads, syncs scores back to Salesforce CRM via REST API, and triggers native Record-Triggered Flows that route hot leads to a dedicated SDR queue and auto-create follow-up tasks, fully closing the loop from data ingestion to CRM workflow execution.

100 leads scored Hot / Warm / Cold routing Bidirectional SFDC sync Zero manual steps

Career OS Mar 2026

Featured View on GitHub →

Python Claude REST APIs GitHub Actions Webhooks

A six-layer AI workflow that converts live job-market data into structured research, skill-gap analysis, sprint plans, and human-reviewed outreach drafts on an automated cadence.

Six-layer workflow Live job-market data Structured research Human-reviewed drafts Automated cadence

Agent Telegram Notifier

View on GitHub → Read the article →

TypeScript Codex Claude Code Telegram Local Hooks

A local-first notification layer for Codex and Claude Code runs. Hooks detect completion, failure, and needs-input states, redact common secret patterns, preserve desktop notifications, and send structured Telegram updates so long-running agent work can be monitored away from the terminal.

Codex + Claude hooks Secret redaction Local-first Dry-run test path

PodBot: AI Podcast Discovery Apr 2026

View the app → GitHub →

React TypeScript Supabase Spotify API OAuth 2.0 TanStack Query Tailwind CSS

Full-stack AI recommendation app that connects to Spotify via OAuth 2.0, analyzes a user's podcast listening history and completion patterns, and surfaces personalized episode picks, not by category tags, but by actual behavioral signal. Taste profile is built from play rates, skip patterns, and listening cadence, then fed into an AI layer that ranks and explains each recommendation. Users can save any pick directly to a Spotify playlist. Backend runs on Supabase edge functions with PostgreSQL persistence.

OAuth 2.0 + token refresh Behavioral taste profiling AI-ranked recommendations

Experience

GTM Systems Analyst Intern

Together AI

Apr 2026 - Present · San Francisco Bay Area

Shipped a Salesforce-embedded production AI answer engine spanning six live data lanes, including Salesforce, product usage, Gong calls, and buying signals; rebuilt its inference path to cut compound-query latency from 165 seconds to 6.3 seconds, a 26x speedup.
Built an adversarial agent-evaluation system with synthetic questions and an independent LLM judge; used paired statistical testing to select the production model and shipped a grounding selector that improved answer accuracy by 7.5 percentage points.
Engineered deterministic grounding and security controls, then automated Gong call insights, intent routing, closed-lost re-engagement, and lead scoring across Clay, Slack, Hex, Snowflake, and Salesforce.

Campus Ambassador

Clay

Jun 2026 - Present

Represent Clay, the GTM data enrichment platform, on the UW-Madison campus

Growth

Roger (YC S24)

Dec 2025 - Mar 2026

Ran outbound growth experiments for a YC-backed AI SDR platform
Built prospect lists, launched campaigns across LinkedIn and email, tracked performance metrics
Created monthly content (TikTok, LinkedIn video) demonstrating product value

GTM Engineer

College and University Healthcare Education Consortium (CAUHEC)

Sep 2025 - Present · Part-time · Remote

Built and own CAUHEC's outbound GTM infrastructure end to end, from lead sourcing and enrichment through deliverability and funnel reporting, as a one-person system
Designed AI-native sourcing and enrichment pipelines in Python, using local LLMs and public data to produce verified, high-intent contacts at scale without paid data vendors
Treated deliverability as an engineering problem, maintained bounce below 5%, and protected sender reputation across a multi-mailbox, multi-persona campaign system
Automated funnel and conversion reporting so growth decisions run on live data, and partner with the founder on segmentation and go-to-market strategy

Data Analytics Intern

Community Action Partnership of Ramsey & Washington Counties

Jul 2025 - Aug 2025

Built dashboards tracking 100+ active community partnerships, filterable by engagement tier, department, and staff lead
Analyzed historical data to surface engagement gaps and regional disparities
Presented findings to agency staff to improve outreach equity and strategy

Earlier leadership: Website Director, TEDxUWMadison (Oct 2024 - Aug 2025)

Education & Credentials

University of Wisconsin-Madison

B.S. Computer Science

Sep 2023 - May 2027

Chi Psi Fraternity · Informatics Skunkworks Machine Learning · Data Science Club

Agentic AI DeepLearning.AI · May 2026
Revenue Operations HubSpot Academy · Oct 2025
Salesforce Admin Fundamentals Trailhead
Clay GTM Cohort Clay

Skills

I combine technical depth with business fluency. Every system I build ties to real revenue outcomes.

AI Engineering

AI Agents LLM Application Development Evals RAG / Retrieval MCP Grounding Guardrails Prompt Engineering Claude Code

GTM Systems

Salesforce SOQL Salesforce Flows Clay HubSpot Apollo Gong Lead Scoring Intent Signals Data Enrichment

Automation & Integration

REST APIs Webhooks OAuth GitHub Actions Slack Workflows Zapier Make

Engineering & Data

Python JavaScript / TypeScript SQL React / Next.js Node.js Supabase Vercel Redis Snowflake Hex dbt Census Tableau

recent jams

recent flicks

Loading…

Let's connect.

If you are hiring for GTM Engineering or Applied AI, or building production systems across AI, CRM, data, and automation, I would be glad to compare notes.

Email LinkedIn GitHub Writing