tanveer@local

run it local

Run a free, private
ChatGPT-like assistant on
your laptop.

$240

ChatGPT Plus / year

Your data on their servers

$0
What I'm about to show you
Your data stays on your machine

whoami

Two brothers. Two playbooks.

AI + Digital Marketing. Singapore + Pakistan.

TR

Tanveer Riaz

VP, Technology Transformation
Global bank, Singapore

20+ years in banking across Asia, Middle East and Africa. By day: AI transformation in transaction banking. By night: building AI tools and shipping products locally.

Community session · personal capacity · not representing my employer

AI Builder Transaction Banking Singapore

linkedin.com/in/tanveerriaz

Connect on LinkedIn

FR

Faizan Riaz

Co-Founder & Director
Forward Digital Consultancy

13+ years in digital marketing - 300+ brands across 17 industries. Tonight he shows how local AI fits marketing: ad copy, campaign research, influencer outreach, and scaling what you build.

Digital Marketing Campaigns Growth

linkedin.com/in/faizanrz

Connect on WhatsApp

agenda.md

tonight's roadmap

Eight things.
All live. All local.

Install Three Interfaces Models Excel RAG Vision Web Search @agent Scheduled Jobs

Then Faizan takes over - local AI meets real marketing: campaigns, content, and growth playbooks.

step 1 - the engine

ollama.com

One install.
One command.

Ollama is the engine. It downloads and runs AI models directly on your machine. No account, no API key, no subscription.

$ ollama pull gemma4:12b-it-qat

LIVE DEMO - install + first run in Terminal

step 2 - the interface

from terminal to product

Same brain.
A face you already know.

AnythingLLM wraps Ollama in a ChatGPT-style interface. Same model, same privacy, but now it looks like a product anyone can use.

LIVE DEMO - AnythingLLM vs Terminal side by side

three ways in

same engine, three doors

Terminal. App.
AnythingLLM.

One local model stack - pick the interface that fits how you work.

⌨️

Terminal

ollama run, scripts, power users. Full control, zero UI overhead.

🦙

Ollama App

Native desktop chat. Download models, switch models, chat in one window.

💬
AnythingLLM
RAG, agents, scheduled jobs. ChatGPT-style product on top of Ollama.

models - pick your tradeoff

switch models in one click

Speed vs reliability.

Different tasks need different models. Swap in Terminal, Ollama app, or AnythingLLM - same pull, same privacy.

gemma4:12b-it-qat

⚡ Speed-first

Quantized for fast replies on a laptop. Great for chat, quick summaries, live demos.

qwen3

🎯 Reliability-first

Stronger reasoning and tool use. Better when accuracy matters more than seconds.

$ ollama pull qwen3 # switch anytime

LIVE DEMO - switch models side by side

demo - excel rag

use case 01 - prompt 3

Your spreadsheet.
Your data. Locally.

Real estate sales sheet - unit prices, commissions, buyer notes. Upload to AnythingLLM and ask: "Which units sold above asking?" The Excel never leaves your machine. No cloud. No training.

Private data Excel / CSV RAG Zero cloud upload

LIVE DEMO - real estate sales sheet Q&A

privacy.md

🔒 why this matters for your money

Your financial questions
stay on YOUR machine.

💻

Local Inference

Model runs on your hardware. No cloud server ever sees your prompt.

🔒

No Training

Nobody is training on your tax questions, property analysis, or portfolio data.

🚫

No Account

No login. No email. No tracking. Download, run, done.

I'll show you the privacy screen inside AnythingLLM itself. Don't take my word for it.

demo - vision

use case 02

Show it a photo.
It sees. Locally.

Upload a receipt, a business card, a product screenshot. The model reads and analyzes images without sending them anywhere.

LIVE DEMO - image upload + analysis

Note: Vision needs a multimodal model (e.g. llava, gemma with vision). If the current model can't see images, switch to a vision-capable pull - same local stack, one command.

demo - @agent

prompt 1 - live now

@agent Wali Khan.
Right now.

Ask AnythingLLM's agent to scrape Wali Khan's YouTube channel and summarize the latest videos. Web search + local reasoning - triggered on demand, not on a schedule.

Prompt: @agent scrape Wali Khan's YouTube. Summarize latest 2 videos.

Mode: On-demand | @agent in chat

Model: gemma4:12b-it-qat (local)

LIVE DEMO - @agent against Wali's channel now

demo - scheduled agent

set it and forget it

Set it once.
It works while you sleep.

Create a scheduled agent in AnythingLLM - daily web search, scrape, summarize, deliver. No API key. No subscription. Show the job config, then the last successful run.

Task: Scrape Wali Khan's YouTube. Summarize latest 2 videos.

Schedule: Daily 7:00 AM

Last run: Completed | 30.4s

Model: gemma4:12b-it-qat (local)

3 bullet points per video. Titles, dates, URLs. All local.

LIVE DEMO - create job + show last successful run

recap.md

what you just saw

One laptop.
Eight capabilities.
$0.

Excel RAG Vision Web Search @agent Scheduled Jobs 100% Local 100% Free

Install steps and every tool I used tonight will land in your inbox after this session.

handoff

part two

I showed you the engine.
Now Faizan shows you
the marketing playbook.

How local AI powers campaigns - ad copy, audience research, influencer workflows, and turning builders into brands. Real marketing, real results.

tanveer@local - wifi: off

close

Still no cloud.
Still $0/month.

Everything you saw tonight runs on a laptop you can buy today.

In the chat right now: what's the first thing you'd use this for?

Run a free, privateChatGPT-like assistant onyour laptop.

Two brothers. Two playbooks.

Tanveer Riaz

Faizan Riaz

Eight things.All live. All local.

One install.One command.

Same brain.A face you already know.

Terminal. App.AnythingLLM.

Speed vs reliability.

Your spreadsheet.Your data. Locally.

Your financial questionsstay on YOUR machine.

Show it a photo.It sees. Locally.

@agent Wali Khan.Right now.

Set it once.It works while you sleep.

One laptop.Eight capabilities.$0.

I showed you the engine.Now Faizan shows youthe marketing playbook.

Still no cloud.Still $0/month.