TA.

CAREER PATH / HANOI

Thúy Anh.

MARKETING · DATA · AI · TRAINING

From manual content operations to systemizing processes with data and AI.
I design processes, train teams, and use AI to do more with less.

Direction: researching & applying AI more broadly to business operations, not just content production.

  1. 1Stamp illustration: coordinating a nationwide competition

    Coordinating a nationwide competition

  2. 2Stamp illustration: building a channel from zero

    Building a channel from zero

  3. 3Stamp illustration: turning insight into a process

    Turning insight into a process

  4. 4Stamp illustration: making films without a crew

    Making films without a crew

  5. 5Stamp illustration: building a storefront without code

    Building a storefront without code

Production cost — AI instead of a crew
Peak monthly revenue
People trained on the AI process
AI Application
Digiphar Pharmaceutical
Content production
2024 – 2026
What I carried forward

Turning a manual process into steps that can be taught to someone else — that's the approach I carry into every project since.

Chapter 01AI Application
Google AI StudioVeo 3MiniMaxSync.soChatGPT / Claude
Stamp illustration: AI application

No crew, still a finished film.

Applying AI to produce media assets, standardizing the process, and training the team.

An AI process that can be handed offFrom script to visuals, voice, and lip-sync; documented and standardized, with ~15–20 people trained directly. Outsourced production cost dropped to near zero (AI tool costs counted separately).
Real result — aired

A TVC built with this exact process

A TVC for a brand I worked on. New packaging made the entire old footage library obsolete; some outdoor scenes also couldn't be shot for real within budget. AI handled ~80% of this TVC's visual production.

Remaining segmentsReal footage, not AI-edited — except the two segments above (0:04–0:08 and 0:09–0:10, can swap).
Production pipeline — from script to final asset
Script Character — still image Google AI Studio Character — motion Veo 3 Voice MiniMax Lip-sync Sync.so Final asset
Problem

A packaging redesign made the entire old footage library (shot at real crew-hire cost) obsolete overnight. At the same time, the brand needed hard, expensive shots that the budget couldn't cover with real filming.

Contribution

Self-researched · personally produced all the media assets · standardized the process & documentation · trained ~15–20 people.

Process
  • Research & tool selection: tested and settled on an AI tool for each step — still image, motion, voice, lip-sync (Google AI Studio · Veo 3 · MiniMax · Sync.so)
  • Rebuilding the assets: ran the pipeline script → character → voice → lip-sync → final asset; hard scenes built 100% with AI, usable real footage edited with AI directly on the original frame (swapping old packaging for new)
  • Packaging the process: standardized into a repeatable pipeline, splitting work between AI and the design team by scene difficulty
  • Documentation: wrote a structured prompt formula (setting → character → action broken down by second) plus step-by-step guides
  • Training: personally coached ~15–20 people, with hands-on exercises
— 014 —

Training materials, the prompt formula, and before/after examples available on request in conversation.

Performance
Digiphar Pharmaceutical
Pharma & medical devices
2023 – 2026
What I carried forward

The habit of starting from real observation — comments, data, behavior — before making any decision

Chapter 02Insight Mining
Stamp illustration: insight mining

Mining insight through a standardized process.

From real data to hypothesis, validation, and a content plan.

100–300M VND / insight / month personal revenue from each insight I mined directly · peak month ~500M VND
Insight-mining process — from observation to content plan
1 Observation 2 Hypothesis 3 Validation 4 Final Insight 5 Big Idea 6 Resource prep
  1. Observation
  2. Hypothesis
  3. Validation
  4. Final Insight
  5. Big Idea
  6. Resource prep
Operations report

A Facebook Ads measurement dashboard I built myself with AI — tracks cost/revenue, ROAS and the conversion funnel by campaign, with automatic diagnosis and recommended actions.

View the full report — multiple campaigns, daily trend, conversion funnel →

Screenshots of the full report (sample data) — view the live version at baocao.thuyanh.site →

Overview for this period and daily cost/ROAS trend
Overview & daily cost/ROAS trend
Conversion funnel and top of the campaign performance table
Conversion funnel & campaign table
Breakdown by platform and audience
Platform & audience breakdown
Automated insights and recommendations
Automated insights
Problem

The brand I supported faced a direct competitor: a nearly identical formula (99%) at a lower price, pushing customers to compare price over value. Needed 1–2 insights strong enough to shift perception and escape the price war.

Contribution

Marketing Performance Lead · mined & validated insight · standardized the process & documentation · monitored data and proposed fixes · led & trained the team.

Process
  • Mining & validating insight: analyzed 100+ comments on TikTok/YouTube, "spied" on Facebook Ads Library, researched competitors & the market; ran the Observation → Hypothesis → Validation loop before locking in 1–2 insights
  • Standardizing into a process: built an end-to-end process for Facebook Messenger Ads, written up as a step-by-step guide others could run
  • Monitoring operations: analyzed campaign & budget data, spotted anomalies → proposed concrete fixes
  • Leading & training: coordinated the team mining insight through the process, managed and trained staff — each insight I mined directly brought in 100–300M VND/month
— 041 —
Coordination & Partnerships
Marketing Hive
DAV Marketers Club · nationwide
2021 – 2022
What I carried forward

This was the starting point — where I learned to stand between many parties to keep a program running smoothly.

Chapter 03Event Coordination
Stamp illustration: event coordination

Coordinating a nationwide competition.

The single point of contact between contestants, sponsors, and schools — keeping internal operations running smoothly.

Hanoi → nationwide the scale of season 2, which I helped coordinate; single point of contact between contestants from many schools, sponsors, and the organizing committee
Attendees at the Marketing Hive event in the hall
Marketing Hive — event photos
Team photo at the Marketing Hive event
A moment with the team
Problem

Marketing Hive season 2 (DAV Marketers Club, Diplomatic Academy of Vietnam): the competition scaled from Hanoi to nationwide — more schools, more contestants and sponsors, and a heavier coordination load. Needed a single point of contact to keep the organizing team running and coordinate everything centrally.

Contribution

Head / Deputy Head of HR · recruited & assigned the organizing team · point of contact for contestants, sponsors & schools · coordinated internal operations.

Process
  • Building the team: recruiting – assigning – onboarding organizing-committee members
  • Multi-party point of contact: the central communication channel between contestants, sponsors, and school representatives
  • Sponsor outreach: worked with the specialist team to reach out to and work with sponsors
  • Internal operations: handled paperwork, scheduling, and the program's internal activities
— 066 —
Built with AI
Freelance work
WooCommerce
Claude · VS Code · AdsPower
What I carried forward

Not knowing how to code is no longer a barrier to building a complete digital product yourself.

Chapter 04Built with AI
Stamp illustration: building a site with AI

Not one line of code, still a working storefront.

2 complete WooCommerce sites — hosting, domain, products, and promotions all handled solo.

2 websites WooCommerce, live and running, delivered to the client · not one line of code written by hand
Causse furniture website noithatlathanh.comFurniture La Hoa Shop website thuyhoa246.comFresh flower shop
Problem

Took on a freelance web-building job with zero coding knowledge — no back-end, no front-end. Used AI for anything that needed code, delivered 2 e-commerce websites.

Contribution

Directed AI to handle code & content · managed hosting/domain solo · built out WooCommerce end to end · delivered 2 websites.

Process
  • Code: Claude edited CSS/theme code directly, going beyond the page builder's limits
  • Content: Claude wrote product descriptions, blog posts, and basic SEO
  • Infrastructure: AdsPower kept hosting, domain, and WordPress login sessions stable
  • WooCommerce: catalog, cart, contact page, promotions
— 028 —
Content & Community
Laluna
Stationery · Top-2 e-commerce brand
2023
What I carried forward

Starting from viewer insight, using content to test hypotheses, then focusing resources on the products and content directions that showed real signal.

Chapter 05Content & Community
Stamp illustration: TikTok content

Growing a channel from 0 to 100K followers

Building a student community through minigames and using content to test product appeal.

0 → ~100K followers gained while running the channel (2023)

Top-performing videos on the channel

View counts at the time of the screenshot.

Minigames · driving engagement
924.7K views
635.4K views
723.9K views
Product content · testing appeal
1.3M views
1.1M views
625.9K views
349.2K views
136.7K views
Problem

Build a TikTok presence and community for Laluna, a stationery brand, with students as the core audience — while also finding which products had the strongest appeal to focus content on.

Contribution

Content Creator · built the minigame strategy · researched product trends · produced content to test hypotheses · analyzed feedback and optimized content.

Channel-building strategy
  • Nailing the insight: tapped into students' love of receiving gifts and discovering school supplies; chose products and a presentation style that felt close to this audience.
  • Minigames for engagement: built giveaway content to encourage comments, likes and shares, drawing viewers in and growing the channel's community.
  • Forming product hypotheses: researched trending products, predicting which designs or product groups might appeal to students.
  • Using content to validate: produced test content for each hypothesis, tracking views and engagement to gauge real interest. A viral piece of content was the signal for picking a "winning" product for content appeal.
  • Doubling down on what worked: kept developing more content angles around the products that got attention, tracking performance and adjusting to keep the channel's appeal up.
— 055 —