AI Agents · Workflow Automation · LLM Integrations

JAI PUNDE.

AI Automation Engineer / Mumbai, India

Automation engineer who ships AI that keeps working after everyone logs off. 50+ production n8n workflows, WhatsApp AI agents, RAG pipelines and CRM integrations, saving US and international clients 30+ hours of manual work every week, with human handoff built in.

View projects →RÉSUMÉEMAILLINKEDINGITHUB
Jai Punde adjusting his sunglasses on a sunny street
Fig. 01 — Jai PundeMumbai --:-- IST

Workflows

n8n systems that survive production

Agents

AI that talks, books and hands off

Integrations

CRMs, APIs and webhooks, wired together

01

ABOUT

I automate the repetitive parts of a business, and make sure a human is one tap away when the AI isn't sure.

I'm an AI automation engineer from Mumbai, finishing a B.E. in Computer Science. I got into automation in 2024 as an intern at Make My Vid, wiring Reddit, LinkedIn, YouTube, HeyGen and Slack into AI pipelines that published content on their own.

Since then I've built recruiting automation at GRR BAOW, and today I build AI agents for US businesses at Closer Control CRM: WhatsApp sales agents, call note-takers, appointment setters and support bots, mostly on n8n and GoHighLevel.

I care about the unglamorous parts. Debouncing people who send five messages in a row. Retries on flaky APIs. Memory that doesn't balloon. Knowing what every run costs. If it can't run unattended at 3am, it isn't done.

Off the clock: motorbikes, monsoon hills and old forts.

0+production workflows
0+hours saved every week
0US clients onboarded
~0%support issues solved by AI
02

PROJECTS

Systems I designed and shipped, from WhatsApp agents to a full-stack SaaS. Each card is a live sketch of what the system does. Hover or tap for the brief.

01

WhatsApp AI Sales Agent

Hover to read ↗

The brief

A WhatsApp sales rep that reads everything, books calls and knows when to hand off.

  • Understands text, voice notes, images, PDFs and CSVs
  • Buffers rapid messages in Redis so it answers once, not five times
  • Checks slots, books, reschedules and cancels through GoHighLevel tools
n8nWAHARedisPostgreSQLGeminiOpenAI

5 input types · 7 agent tools · human handoff

02

AI Call Summarizer

Hover to read ↗

The brief

Every sales call becomes a lead score, a CRM note and a coaching tip.

  • Speaker-labelled transcripts with talk time per speaker
  • Strict JSON output: score, objections, next steps, coaching
  • Reads real audio length from the file header to skip short calls
n8nOpenAI transcriptionStructured outputsGoHighLevel APIGHL Marketplace

lead score 1–10 · hot / warm / cold

03

URL to Media SaaS

Hover to read ↗

The brief

Paste a website, get on-brand images and videos.

  • AI-written brand document from the website and brochure (RAG)
  • Text→image, image→image, text→video and image→video, logo on or off
  • Async jobs with callbacks, timeouts and error logging
Next.jsSupabasepgvectorn8nClaude APIOpenAI

4 generation modes · brand-grounded

04

Automated Lead Revival

Hover to read ↗

The brief

Wakes up cold leads with local-presence SMS and sorts every reply.

  • Local-presence SMS drips that run around the clock
  • AI sorts every reply into 15 disposition stages
  • Interested leads go straight to the AI appointment setter
n8n AI AgentGoHighLevelGHL Marketplace AppOAuth2SMS

~20% reply rate · 15 dispositions

05

Stock Report Insights

Hover to read ↗

The brief

Telegram PDF reports in, personalised client emails out.

  • Telethon listener on stock news channels
  • PyMuPDF text extraction from new PDF reports
  • Webhook hand-off from Python to n8n
PythonTelethonPyMuPDFn8nWebhooksLLM APIs

PDF → insights → inbox

06

AI Support Bot

Hover to read ↗

The brief

Resolves most support issues on its own and escalates the rest with context.

  • Answers, triages and opens GoHighLevel tickets
  • Resolves about 70% of issues end to end
  • Escalates the rest with full context to a human
n8nLLM agentsGoHighLevelTicketing

~70% resolved without a human

WATCH THE RUN →
07

AI Appointment Setter

n8n + GHL Marketplace app that answers new leads, qualifies them and books calls.

Agent
08

ClickUp Task Analyzer

Flags delays, creates dependent tasks when work is blocked, alerts clients on missed deadlines.

Ops
09

Real-Estate Lead Gen

Scrapes listings, filters quality prospects and pushes them into GoHighLevel.

Scraping
10

KPI Dashboards

Next.js + Supabase dashboards for active leads, lead sources and deliverability per account.

Full-stack
11

LinkedIn Recruiting Outreach

GetSales and Dripify sequences with AI replies grounded in job descriptions, plus recruiter alerts.

Agent
12

Loxo CRM Sync

Candidate sync between GetSales and Loxo with live tracking of applications and recruiter bonuses.

Integration
13

RAG Document Pipelines

Embeddings and vector search that feed dynamic context into automated content generation.

RAG
14

Multi-platform Content Engine

Reddit, LinkedIn, YouTube, HeyGen and Slack APIs wired to AI models to publish content.

Content
15

GHL Onboarding at Scale

150 US clients: A2P 10DLC, pipelines, booking and follow-up automations, ads integrations, lead data prep.

CRM
03

EXPERIENCE

  1. AI Automation Engineer · Closer Control CRM Now

    Jan 2026 — Present

    San Francisco, CA, USA · Remote

    Building AI agents and automations for US businesses on n8n and GoHighLevel.

    • Built an n8n AI note taker that transcribes lead calls, scores each lead with an LLM and delivers structured call summaries.
    • Built an AI appointment setter (n8n + GHL Marketplace app) that responds to new leads, qualifies them and books calls.
    • Developed an AI support bot that creates GHL tickets and resolves about 70% of issues without a human, escalating the rest.
    • Designed WhatsApp AI agents for industries like car dealerships that send contextual messages and images based on customer intent.
    n8nGoHighLevelOpenAIWhatsAppNext.jsSupabase

  2. AI Automation Engineer · GRR BAOW

    Dec 2024 — Jan 2026

    Pune, India

    Recruiting automation, CRM sync and RAG pipelines.

    • Automated LinkedIn candidate outreach sequences in n8n (GetSales, Dripify) with AI replies based on job descriptions and internal notes, plus recruiter alerts for complex queries.
    • Integrated GetSales with Loxo CRM for candidate sync; built live tracking of job applications and recruiter bonuses.
    • Developed RAG document pipelines (embeddings, vector search) for dynamic context and automated content generation.
    • Built authenticated apps with Next.js frontends, Supabase backends and private storage for data confidentiality and user isolation.
    n8nGetSalesLoxoRAGNext.jsSupabase

  3. AI Automation Intern · Make My Vid

    Aug 2024 — Dec 2024

    Mumbai, India

    Where the first workflows shipped.

    • Built fully automated AI pipelines in n8n for lead generation, content creation and client engagement.
    • Integrated Reddit, LinkedIn, YouTube, HeyGen and Slack APIs with AI models to publish content across platforms.
    n8nHeyGenSlackYouTube API
04

TECH STACK

Workflow automation

n8nAI AgentsWebhooksDatabase TriggersError HandlingMake.comZapier

AI & LLMs

OpenAI APIClaude APIGeminiLangChainRAGEmbeddingsVector SearchPrompt EngineeringStructured Outputs

APIs & integrations

REST APIsOAuth2Web ScrapingWhatsApp (WAHA)SlackClickUpLinkedInHeyGenBlotato

CRM

GoHighLevelGHL APIMarketplace AppsA2P 10DLCSalesforceHubSpotREsimpliLoxoGetSales

Development & data

PythonJavaScriptSQLPostgreSQLSupabaseRedisNext.jsTailwind CSSGitClaude CodeMCP
05

HOW I BUILD

The habits behind the workflows. None of this is glamorous, all of it is why things keep running.

01

Debounce the humans

People send five messages in a row. Buffer them, then answer once, like a person would.

02

Every agent has an off switch

One tag turns the AI off and a human takes over. Handoff is a feature, not a failure.

03

Memory gets compacted

Long conversations are summarized so context stays small, cheap and on-topic.

04

Know what every run costs

Token usage is measured per execution, so pricing is a formula, not a guess.

05

Structured outputs or it didn't happen

LLMs return strict JSON, so the next step never has to guess what came back.

06

Fail loudly, retry quietly

Retries for flaky APIs, error workflows for everything else, and logs that say what broke.

06

CONTACT

LET'S AUTOMATE SOMETHING.

Got a process that eats hours every week? Send it over. I'll tell you honestly whether it should be automated, and how I'd build it.

Based in
Mumbai, India
Timezone
IST · UTC+5:30 · --:--
Email me →

Channels

Emailjaipunde@gmail.com →LinkedInin/jai-punde ↗GitHubgithub.com/jaipunde ↗RésuméView PDF ↗The RunStory mode →