Fenil Parekh

GTM Engineer · AI agents + revenue systems

location: San Francisco Bay Area
local: — · — · PST
since: 2017 · nine years in
now: Marketing Boutique INC · consulting

currently shipping agentic GTM infra · HR Tech B2B SaaS + $30M+ CPG

open to

ad-hoc audits short engagements agentic GTM partnerships

◷ Nine years on the axis

I'm a GTM engineer based in San Francisco Bay Area. I build internal AI agents and go-to-market AI agents that run inbound, outbound, paid, and email for AI-native B2B SaaS. Internal agents help the GTM team — AEs, SDRs, BDRs, RevOps — work pipeline. Go-to-market agents run a surface directly under human supervision. The two classes have different design constraints, supervision models, and failure modes, and the distinction is most of the work.

The work sits at the intersection of AI agents and revenue systems. The next decade of GTM is being rewritten there — answer engines reshaping discovery, agents handling research and personalization, the four surfaces all becoming engineering surfaces. I'm trying to be one of the small number of people who sits credibly at that intersection, with the code and the case studies to show for it.

⌬ The intersection I work at

How I work

six principles

Architecture beats tools.
architecture

Tools change every quarter. The architecture of how data, intelligence, and activation fit together does not. Encyclopedia content here uses category-level language; tool specifics live on /stack.
Failure modes are the credibility signal.
reliability

Every case study names what broke. Every encyclopedia page names the constraints — deliverability, crawl budget, brand-voice drift, attribution gaps — that decide whether a system survives production.
Numbers come from real builds or they don't appear.
rigor

No projected pipeline, no aggregated CAC, no "typical open rate." Just the runtime, the token spend, the dedup pass rate, the pages indexed — from a specific build, attributed to a specific case study.
TDD on agents is non-negotiable.
quality

Pytest unit tests on every agent node, integration tests on external APIs with VCR-recorded fixtures, LLM-as-judge eval sets that gate prompt regressions in CI.
Internal vs. go-to-market is a real distinction.
taxonomy

Most "agentic GTM" content blurs the two. Internal agents have implicit supervision (a human reviews every output). Go-to-market agents have explicit supervision (HITL gate, brand-voice validator, factuality check).
Built and broken in the open.
transparency

I write what I ship and what breaks. The newsletter, the library, the encyclopedia — they exist to compound a body of work that survives review.

Work history

four entries

2025 — Present Marketing Boutique INC current

Lead GTM Engineer (consulting)

Deployed zero-to-one agentic GTM infrastructure across HR Tech B2B SaaS startups and e-commerce CPG brands at $30M+ Amazon revenue. Built production reasoning loops in Python with schema-validated JSON outputs and human-in-the-loop approval gates. Strict TDD: pytest per agent node, integration tests with VCR fixtures, LLM-as-judge evals gating prompt regressions in CI.
- agentic GTM
- $30M+ CPG
- HR Tech B2B SaaS
stack
- Python · daily
- TDD · pytest
- HITL gates
- LLM-as-judge evals
2023 — 2025 Ezyhire archived

Founding GTM Engineer (ABM & Demand Gen)

Built a zero-touch outbound engine: waterfall enrichment across multiple data providers, RAG-grounded LLM prompts generating hyper-personalized first-touches, factuality validation before every send. Wired intent signals into routing logic surfacing high-intent accounts to AEs in Slack within minutes. Provisioned outbound infrastructure with warm-up, domain rotation, and algorithmic load-balancing.
- zero-touch outbound
- <5min intent→AE
- 80%+ enrichment hit
stack
- waterfall enrichment
- RAG-grounded LLMs
- factuality validation
- warm-up + rotation
2021 — 2022 CA-One Tech Cloud archived

Demand Generation Manager

40% pipeline-velocity lift via CRM nurture sequences with behavioral segmentation. Built marketing-to-Salesforce sync with JSON webhooks pulling speed-to-lead from 24h to <5min. Topic-cluster SEO strategy delivered 3× organic inbound. Paid across Google/Meta/LinkedIn/Reddit — 30% CAC reduction at 4× ROAS.
- +40% pipeline velocity
- 3× organic inbound
- −30% CAC
- 4× ROAS
stack
- CRM nurture
- JSON webhooks
- topic-cluster SEO
- paid · multi-channel
2017 — 2021 Embtel Solutions archived

Growth Marketing Manager

Full-funnel digital campaigns for SMB clients — landing pages, ad network configuration, attribution — 25% YoY revenue growth. Configured tracking schemas across GTM and GA, improving ROI 30%. CRO-optimized landing pages with A/B testing: 40% lift visitor-to-lead. Parallel SEO and PPC programs that delivered 3× MQL.
- +25% YoY revenue
- +30% ROI
- +40% v→l lift
- 3× MQL
stack
- SMB digital
- attribution
- A/B testing
- SEO + PPC

Skills + tech stack

nine layers

Tools I've shipped real builds on, grouped by layer. The exhaustive "what I use and why" page is at /stack.

Primary language + engineering 07: Python (daily)
TDD with pytest
VCR integration tests
LLM-as-judge evals
GitHub Actions CI
Docker
REST/GraphQL APIs
Agent frameworks + LLM 07: LangChain
LangGraph
Claude Agent SDK
Anthropic Managed Agents
MCP
RAG retrieval
schema-validated outputs
Deployment + observability 04: AWS Lambda + EventBridge
Modal
Render
LangSmith
Workflow orchestration 03: Make.com
n8n
Zapier
Data + enrichment 07: Clay
Apollo.io
Prospeo
6sense
Factors.ai
Apify
LinkedIn Sales Navigator
Intelligence 05: OpenAI API
Anthropic API
custom retrievers
brand-voice validators
LLM-as-judge evaluators
Activation 07: Smartlead
HeyReach
LinkedIn Ads
Meta Ads
Google Ads
Outreach.io
Salesloft
CRM 03: Salesforce (REST API)
HubSpot
Attio
Analytics 05: GA4
Hockeystack
Looker Studio
SEMrush
Ahrefs

Education

M.S. Computer Science

ITU, San Jose

2018 · graduated · California, USA

Résumé

fenil-parekh-resume.pdf

PDF · ~2 pages · updated 2026-05

Download →

Contact

Or use the contact form.