--- title: "GTM and RevOps systems" description: "Ayush Rai on GTM engineering and revenue operations: a live revenue-role job platform, a public-source relationship pipeline, and a synthetic scoring demo with a human approval gate." canonical: "https://www.ayushrai.site/gtm-revops/" last-updated: "2026-10-18" --- # GTM and RevOps systems Source: https://www.ayushrai.site/gtm-revops/ Markdown: https://www.ayushrai.site/gtm-revops.md Ayush Rai on GTM engineering and revenue operations: a live revenue-role job platform, a public-source relationship pipeline, and a synthetic scoring demo with a human approval gate. Evidence level: Mixed: shipped pieces and explicit limits. ## Scope GTM engineering, as I use the term here, is the data and decision layer behind revenue work: finding accounts, resolving duplicates, keeping track of where each fact came from, scoring fit in a way a person can inspect, and keeping a human between a model's output and anything that leaves the building. I have not worked in a RevOps or sales team, and I have no CRM admin, customer, or pipeline-revenue numbers to show. What follows is software I built and the limits its own repositories state. ## RevenueRoles: a live job platform for revenue-technical roles RevenueRoles finds roles from supported employer career sources, extracts the full description, checks whether a posting is eligible to publish, and classifies it by the work rather than the title. A GTM Engineer is not a Sales Ops role, and the taxonomy reflects that. Each listing links back to the original posting. The stack is Next.js, a Python FastAPI worker, Supabase Postgres and Inngest. The live homepage showed 71 verified roles, last updated 23 September 2026, and a compensation coverage figure of 58% when I checked it on 18 October 2026. Those are the site's own counts on that day. They are not user numbers, and I have no revenue or user figures to publish. - Pipeline: discover, extract, verify eligibility, classify, deduplicate. - Concepts it shows: a clean data layer, duplicate resolution, enrichment quality, and how fresh a signal is. - Known limits: listing freshness, salary extraction that needs unit checks, and merging duplicate postings without losing the apply link. ## Relationship Intelligence Engine: provenance on every field A TypeScript pipeline that discovers technical founders and companies hiring AI and ML engineers from public sources: the Y Combinator public directory, GitHub signals and accelerator portfolio pages. It normalizes records, resolves entities, verifies fields, detects signals, scores and exports. Every important field keeps its value, source, source URL, retrieval time and confidence, and carries one of five statuses: verified, inferred, unverified, conflicting or missing. Unknown values stay empty. The repository says the engine is a research tool and not a lead factory. - Concepts it shows: account discovery, buying signals, enrichment in principle, and data provenance. - Limits: no paid enrichment vendor is wired in, and the audit figures inside the repository are its own notes, not independent checks. ## Revenue Intelligence Engine: a demo on fictional data A demo GTM pipeline. Given an ICP, it loads 42 labeled fictional companies, drops duplicates, scores the rest on a deterministic 100-point rubric, writes structured account research for the top accounts, and holds everything in an approval queue before any export. All companies, contacts and signals in it are synthetic. Enrichment is a mock. It does not send outbound messages, does not write to a live CRM, and has no users or revenue. The number comes from rules, not from a language model; the model is used only for the account research text, with a deterministic fallback when no key is set. - Score components: ICP fit, growth signal, hiring signal, technology fit, GTM need and contactability. - Stack: Python, FastAPI, Next.js, pytest. - Concepts it shows: orchestration, explainable lead scoring, and a human approval gate. ## Supporting work SpiderWeb turns the data export LinkedIn gives you into a searchable map of your own network, with duplicate resolution and clearly labeled rules. It is in development and is not deployed in production. MailU is a local tool that syncs Gmail, categorizes mail, and answers questions over it with search and a local model. It shows email-data and retrieval skills. I have not re-run its tests for this page. ## What I would build next Connect one real enrichment provider behind the Revenue Intelligence Engine's existing interface with rate-limit handling, keeping the synthetic mode. Add a step that pushes approved accounts into a CRM sandbox. Write a short note for each system on why it is built that way, what failed, and what it costs to run. None of this is built yet. ## Evidence - [RevenueRoles (live site)](https://revenueroles.in): The job platform. Counts above are from its homepage on 18 October 2026. - [RevenueRoles repository](https://github.com/ayushrai-hub/RevenueRoles): Web app, API, Python pipeline worker and database migrations. - [Relationship Intelligence Engine repository](https://github.com/ayushrai-hub/SYSTEM-job): Public-source pipeline with field-level provenance. - [Revenue Intelligence Engine repository](https://github.com/ayushrai-hub/REVENUE-INTELLIGENCE-ENGINE): Synthetic-data demo. All data in it is fictional. - [SpiderWeb repository](https://github.com/ayushrai-hub/SpiderWeb): In development. - [MailU repository](https://github.com/ayushrai-hub/MailU): Local Gmail tool. ## Related - [LLM systems in products](/ai-engineering/llm-systems) - [LLM evaluation](/ai-engineering/llm-evaluation) - [Software engineering](/software-engineering) - [GTM engineering](/gtm) - [Contact](/contact) ## Gaps - No experience in a sales, RevOps or CRM-admin role, and no CRM, Clay or Apollo integration in any repository. - No customer, user or revenue figures for any system on this page. - The Revenue Intelligence Engine uses fictional data only. It is not evidence of results on real leads. - The Relationship Intelligence Engine is a personal research tool, and its own audit notes were not independently re-run. Contact: /contact _Unverified award metrics are omitted. Wording matches the public site._