Haesung Lee Sr. BI Analyst · AEG / LA Kings Data · Product · Community

Diving deep into data, surfacing the next wave.

Turning signals into AI products, agent workflows, and communities that move culture forward.

Get in touch See the work GitHub ↗
01 K-Town Night A data insight that became a national story 02 Hesse · AI Video Agent An open-source agent that directs & ships video 03 Customer Segmentation Data science: K-Means clustering to CRM strategy 04 Signal AI builder community + AI newsfeed 05 Gru · AI Fitness Mirror Multimodal 0→1 product, patented
01 · Featured · LA Kings

The NHL's first K-Town Night started as a query.

Fan data showed an underserved demographic hiding in plain sight. I pitched it, built the case, and helped launch K-Town Night, now the franchise's most viral moment in 13 years. The game sold out two years running, and a Koreatown seniors' harmonica class ended up playing the national anthem at the Stanley Cup Playoffs.

Haesung Lee at Crypto.com Arena
Exclusive Interview · The Korea Daily · Feb 2026

Why the LA Kings created K-Town Night instead of Korea Night

On the data-driven strategy behind the event, and why the naming mattered.
As seen in ESPN, Billboard, CBS, The Washington Post, Los Angeles Times
Press index National coverage
ESPN: How a group of Korean harmonica players captivated the world The Washington Post: How Koreatown senior citizens with harmonicas became NHL playoff stars Billboard: L.A. Kings put Koreatown Senior Center harmonica class at center ice Los Angeles Times: How Koreatown seniors have become Kings' playoff good luck charm CBS News: Koreatown seniors win over hockey fans at LA Kings games
2× sold out
Back-to-back seasons
+77% YoY
Specialty merchandise sales
13 years
Franchise's biggest viral moment since
02 · AI video agent · Open source

Hesse: AI Video Director Agent in your pocket.

A phone-operated AI video production pipeline built on Claude Code. One command from my phone starts the loop: a producer skill turns the idea into characters, locations, and a storyboard; Higgsfield generates the shots; renders land in Google Drive to approve or revise, while the Mac runs unattended. Fully open source.

View on GitHub →
Output · generated & assembled end-to-end by Hesse
01 · Phone Claude app · commands + approval gate prompt 02 · Remote Control QR session · push updates session 03 · Hesse Claude Code · producer + visuality skills render 04 · Generate Higgsfield CLI/MCP · image + video link 05 · Review Google Drive · share links approve / revise Idea → Storyboard → Renders → Rough cut
Core production loop · command → orchestrate → generate → approve
Claude Code skills Higgsfield CLI/MCP Phone-operated tmux always-on runtime ffmpeg rough cut
03 · Analytics practice · AEG / LA Kings

Audience intelligence, built for action.

From customer segmentation and behavioral analysis to cultural CRM campaigns, my work turns hidden audience signals into strategies that grow revenue, build community, and create moments people remember.

Customer segments based on demographics and purchasing behavior: K-Means clusters plotted on principal components

Purchase-journey model + k-means segmentation

Modeled the fan purchase journey, then clustered fans to surface the highest-intent lead group for the sales team to target first.

→ +34% YoY on new full-season-membership conversion
Fan Universe dashboard with fan counts, engagement trends, and segmentation filters

LA Kings KPI Dashboard

Partnered with engineering, ticketing, and marketing to define shared success metrics and ship one dashboard, adopted as leadership's single source of truth.

→ Shared metrics + root-cause analysis of conversion
Illustrative visuals for demonstration only, not screenshots of actual organizational data.
Databricks Star schema Medallion (bronze/silver/gold) BigQuery · Fivetran SQL · Python Tableau · Power BI
04 · AI pipeline · Octomatic.ai

Signal: turning AI curiosity into community.

Built an AI-coding community that brought 500+ people into live seminars, then created Signal, a daily AI newsletter that curates the most important updates from Reddit, Hacker News, X, and 150+ sources into one briefing.

Signal Newsletter Pipeline Architecture

Reddit Hacker News X 150+ items / day Ingest Rank Dedupe Daily newsletter 500+ registrations 130+ daily readers Authored the PRD + API contracts
05 · Multimodal product · Head of Product, IPLAB Seoul

Gru: a mirror that sees you in 3D.

A 0→1 AI fitness mirror: hardware, software, content, and app. Multi-camera and sensor arrays capture body data, generate a live 3D avatar, and coach your pose in real time. Filed as a WIPO/USPTO patent (PCT/US2020/061822) and backed by $50K+ in non-dilutive grants from KISED and the City of Seoul.

Capture

Multi-camera / sensor array

Synchronized cameras and sensors capture full-body pose and movement data.

Model

Live 3D avatar

Body data is reconstructed into a real-time 3D avatar of the user.

Coach

Pose coaching & styling

The avatar drives pose correction and personalized outfit recommendations.

WIPO/USPTO PCT/US2020/061822 $50K+ non-dilutive grants +24% usability score Hardware + SW + content + app
WIPO / USPTO · PCT/US2020/061822

Smart mirror system for collecting and visualizing physical appearance information

The multi-camera body-measurement and 3D-avatar coaching system behind Gru is filed as an international PCT application covering capture, avatar generation, and personalized styling recommendations.

View the filing →
Patent figure: mirror display coaching a yoga pose Patent figure: whole-body scan with 3D avatar panel Patent figure: outfit recommendation on the mirror display Patent FIG. 8: system architecture with sensors, processing unit, network, and server Patent FIG. 7: processing unit with CPU, GPU, AI chip, and accelerators
06 · Archive

More builds & experiments.

Live product Listro.menu +
An AI platform that turns a restaurant's existing Google/Yelp photos into studio-quality digital menus (Gemini API), with a "Perfect Pairing" upsell and one-tap review prompts that grew review posts 300%. Live with paying restaurants. Owned end-to-end: PRD, API, Next.js/Supabase. listro.menu →
Live product Tabo.cards +
A digital business-card platform that auto-logs where and when each contact was exchanged, and uses an LLM to instantly draft personalized follow-ups. tabo.cards →
Growth · Com2uS USA A/B testing at 100M+ downloads scale +
Ran store-listing experiments (Google Play Console) that lifted conversion of a 100M+ download title by 7-15%, and drove 50%+ higher conversion with 28× CPI efficiency through data-driven creative strategy across acquisition channels.
Grad research · UCI MS Business Analytics projects +
Selected coursework and research from the MS in Business Analytics: predictive modeling, NLP, and applied ML case studies. Full project list →
07 · About
Currently
RoleSr. BI Analyst · AEG / LA Kings
BaseLos Angeles, CA
BuildingHesse, an open-source AI video agent
StackDatabricks · SQL · Python · Claude Code
Open toAnalytics engineering & AI-native teams

Haesung Lee

I've spent the last five-plus years moving between data, product, and community, and I've stopped treating them as separate jobs. The best work I've done sits at the intersection: a segmentation model that becomes a sold-out night, a sensor array that becomes a coach in your mirror, a feed pipeline that becomes a daily habit for 130 readers.

These days that means bringing AI into the workflow directly: Claude Code, LLM APIs, and shipping small products end to end. I'm most interested in teams building multimodal and foundation models, where the distance between a capability and a community is still being invented.

Emailhaesung.lee.810@gmail.com Site / GitHubleo-hs.github.io LinkedInin/haesung-lee