Prod.

Erkin Mukhamedov

Role

The bridge between marketing and tech.

1,508 video ads · 53 clients · $8.54 per lead

Slate

Source: 2023:05

AI Marketing Engineer, Milpitas, CA.

This site is cut like a video edit. Each clip on the timeline is one chapter, shot in its own look. Scroll, drag the playhead, or use J K L.

I started in marketing, making videos by hand.
Shot, cut, posted. Instagram grew 30%.
The numbers kept going up.

Wide. Campus, golden hour.
Vertical. Student to camera.
Cut on the beat.
Post it. Watch it climb.
CIS recruiting: arrivals.
BTS. They film each other.

+30%Instagram subscribers

Handmade

Company: Shorelight Education
Role: Marketing Associate
Place: Auburn, AL
Source: 2023:05 – 2025:05

Produced and edited social video that grew Instagram subscribers by 30%.
Built a recruiting campaign for students from CIS countries with the recruiting team.

Hand toolsPremiere Pro · CapCut · Photoshop

Every frame cut by hand, one clip at a time.

But everything still ran by hand.
The ideas were bigger than the tools.

+29%

subscriber base

Social media campaigns that grew the subscriber base by 29%.
Events featuring Silicon Valley founders and angel investors.
Launch of the Silkroad Angels Club, connecting Central Eurasian angel investors with Silicon Valley startups.

Silkroad Innovation Hub
2024–2025

Measured

Company: Silkroad Innovation Hub
Role: Digital Marketing Intern
Place: Palo Alto, CA
Source: 2024:05 – 2025:08

ChannelsSocial media · founder and investor events

Good numbers. A small frame.

So with friends, I started a company.
It didn't go that well.
Still, a top Alabama accelerator said yes.

SAKEAI support for international studentsPitch · 2024

Product
Marketing
Sales

Pitched investors.
Tiger Cage 2025, Auburn.

Erkin pitching SAKE at Tiger Cage 2025. — Pitch day

Erkin and a teammate pitching SAKE on stage at Tiger Cage 2025, judges in the foreground. — The team

Founded

Company: SAKE
Role: Founder
Source: 2024:09 – 2025:07

Led product, marketing, and sales for an AI platform that helps universities support international students.

It didn't work. It still changed everything.

I realized I love tech.
But I have no experience.
So I taught myself to code.
The AI boom had just begun. I went all in.
A year of small marketing automations.

Low-poly game scene: a young man, seen from behind, works on a laptop in a dark room at night.

# learning to code
print("hello, world")
# first small apps, then the AI boom
curiosity = ["AI", "ML"]
# a year of simple marketing automations
for task in marketing:
automate(task)

python learning.py
SyntaxError: invalid syntax
python learning.py
hello, world

~/automations6 small tools

caption-botWrites post captions from a photo and a tone of voice.
post-schedulerQueues a week of social posts from one spreadsheet.
lead-sorterScores new leads and puts the hot ones on top.
ad-variantsTurns one ad idea into ten copy variants to test.
lookscleanAn AI agent that tidies files on a Mac, with a preview first.
outreach-mvpA/B tests SMS and email campaigns against trucking leads.

Learned

Study: Self-taught programming
Focus: AI, ML, marketing automations
Source: 2025:07 – 2026:01

Nights, a laptop, and a lot of error messages.

Then the Bay Area: Call2Action.
First, simple Telegram bots.
Then the platform: brief in, ads out.
From here, the agent does the cutting.

Erkin Mukhamedov at the Palace of Fine Arts, San Francisco.

AI Creative EngineerBay Areasince Dec 2025

Telegram bots
The platform
1,508 video ads

San Francisco Bay Area

Ad botPlaceholder data

/start
Send me a brief and I'll make the ads.
Brief: CDL-A drivers, home weekly. Two vertical ads.
On it: casting, script, render.
Ad 1 of 2 is ready for review.

One bot per job:

Ads
Landing pages
Social posts
Lead research
Reporting
Motion video

The ad agent platform in demo mode: an agent session turning a brief into video ads. Client details replaced with placeholders.

Briefclient brief in
Castingreference casting
Promptsprompt models
RenderSeedanceKling
VoiceElevenLabs
Subtitles + musicpronunciation checks
QAvision-based checks
DeliveredMetaTikTok

Demo ad frame, made-up ad: “Now hiring CDL-A drivers” — Demo · made-up ads

Demo ad frame, made-up ad: “Home every weekend” — Demo · made-up ads

Automated

Company: Call2Action
Role: AI Creative Engineer
Place: Bay Area
Source: 2025:12 – 2026:11

The rest of the platform

Internal AI platform with custom agents for 30+ enterprise clients
Landing pages with Meta Conversions API and TikTok Events tracking
Lead scoring and routing with classifier models
Intent-based prospecting over 12,000 target companies
Stripe-to-Meta budget ledger, daily dashboards, low-balance alerts
Routine agent lookups 2x faster and 3x cheaper

I stopped cutting. I started building the cutter.

Then I scanned every ad for what made it win.
$122k of Meta spend. 660 ads. 14,330 leads.
$8.54 per lead.

Demo ad frame, made-up ad: “New trucks. Real miles.”

Demo ad frame, made-up ad: “Apply in 2 minutes”

Demo ad frame, made-up ad: “More time at home”

Demo ad frame, made-up ad: “Start this month”

default direction <- winners

real drivers · fast apply · warm voice · upbeat music

Demo values. The real analysis covered 660 ads.

Analysed

Company: Call2Action
Role: AI Creative Engineer
Source: 2026:03 – 2026:11

Traced the winning ads back to their creative inputs and set the agent's default creative direction from the best performers.

Find what wins. Make it the default.

And I kept building.
Harnesses, models, and the systems around them.

1,508A brief-to-video-ad agent: Seedance, Kling, ElevenLabs and vision QA. 1,508 ads for 53 clients.
30+End-to-end agent harnesses for 30+ enterprise clients: ads, landing pages, leads and reporting.
27BFine-tuned Qwen3.8-27B with LoRA and QLoRA. Evals first: held-out clients, blind LLM judge.
12kIntent-based prospecting over 12,000 companies, with lead scoring and routing.
3xAgent lookups made 2x faster and 3x cheaper through A/B tests and defect analysis.

M.S. Computer Science, Sofia University, Palo Alto · GPA 4.0 · enrolled
B.S. Business Administration, Marketing, Auburn University · GPA 3.95

Engineered

Where: Call2Action and own projects
Role: AI engineer
Source: 2026:01 – 2026:11

QwenHugging FacevLLMOllama

Under every ad, a prompt. Under every prompt, weights.

Handmade
Measured
Founded
Learned
Automated
Analysed
Engineered

Looking for an AI marketing engineer? Take the résumé, or write to me.

Export

Source: 2026:11

Open to AI marketing engineer roles: growth teams that want ads, landing pages, tracking and reporting built as systems.

I started in marketing, making videos by hand.

Shot, cut, posted. Instagram grew 30%.

The numbers kept going up.

But everything still ran by hand.

The ideas were bigger than the tools.

So with friends, I started a company.

It didn't go that well.

Still, a top Alabama accelerator said yes.

ErkinI realized I love tech.

ErkinBut I have no experience.

So I taught myself to code.

The AI boom had just begun. I went all in.

A year of small marketing automations.

Then the Bay Area: Call2Action.

First, simple Telegram bots.

Then the platform: brief in, ads out.

From here, the agent does the cutting.

>Then I scanned every ad for what made it win.

>$122k of Meta spend. 660 ads. 14,330 leads.

>$8.54 per lead.

And I kept building.

Harnesses, models, and the systems around them.

The bridge between marketing and tech.

Erkin Mukhamedov

Handmade: Shorelight Education, 2023:05 to 2025:05

Measured: Silkroad Innovation Hub, 2024:05 to 2025:08

Founded: SAKE, 2024:09 to 2025:07

Learned: self-taught programming, 2025:07 to 2026:01

Automated: Call2Action, 2025:12 to now

Analysed: scanning ads for what made them win

Engineered: what I build as an AI engineer

Export: résumé and contact