GH KIM

AI Systems & Technology Leadership.

I build the AI systems that move operational work. Natural-language analytics over data warehouses, autonomous agents for ongoing ops, and the pipelines that keep them running. A decade-plus delivering technology and operations programs across healthcare, education, cloud, and advertising, now applied directly to AI systems.

Start the Conversation LinkedIn
HUN KIM
Hun Kim
What I Do

What I build. AI systems for the operating side of companies. Natural-language analytics over data warehouses. Autonomous agents for ongoing ops. AI-augmented ranking, evaluation, and structured-data extraction. The work sits between leadership's questions and the underlying data, turning recurring "I wish I knew X" into something you can ask.

Who it's for. Founders, CEOs, and senior leaders putting AI into the operating layer of their business. Not as a chatbot demo or a marketing add-on. Open to fractional, advisory, or full-time engagements depending on fit.

What I'm not. Not a researcher, a prompt consultant, or a marketing-AI vendor. Not someone who hands you a deck and walks away. I build systems that ship and stay shipped.

Selected Work

The Intelligence Play

Role
Technical Architect & Developer
The Mission
Eliminate latency and noise in high-volume job tracking by building an AI-powered candidate-matching platform.
The Build
Multi-source aggregation engine scraping job boards via RSS and JSON-LD extraction. AI ranking system using the Groq API to grade listings A+ through F against a candidate profile with written reasoning. Automated daily digest pipeline delivering top matches to a dashboard and personalized newsletters.
The Outcome
Zero-infrastructure SaaS prototype with AI-driven relevance scoring; manual review time cut from hours to minutes. Built for white-label deployment across specialized hiring verticals.
Open Source
sheet-grader: the LLM grading pattern from this project, extracted as a reusable Google Apps Script (MIT-licensed).

The Agent Play

Role
Builder & Operator
The Mission
Build a personal AI operating system that handles the recurring shape of a working day (calendar, mail, tasks, weekly briefs, ongoing strategic context) so the day runs on anticipation rather than reactive triage.
The Build
A custom autonomous agent on a scheduled-task architecture, with a long-term memory layer (pgvector), a Telegram channel for proactive push, and over a dozen subsystems covering daily briefs, weekly reports, calendar and mail processing, networking surfacing, persona research, and personal-finance review. Built and operated by one person, in plain English, via AI-assisted tooling.
The Outcome
In daily production use since April 2026. Replaced a half-dozen disconnected AI workflows with a single unified agent. Codename: Hawat. A working proof-of-concept for autonomous personal-ops agent design.

The Automation Play

Role
Strategy & Operations Manager
The Mission
Eliminate manual data entry, standardize global variance management, and put on-demand budget intelligence in the hands of marketing leadership.
The Build
Architected and shipped an automated ETL pipeline (Google Sheets → Python → dbt → data warehouse) holding 100% daily continuity. Designed and deployed an AI-powered natural-language analytics interface over the warehouse, letting leadership query budget data conversationally rather than waiting on analyst cycles.
The Outcome
100% reporting reliability across global teams; manual data entry eliminated; budget questions answered in seconds instead of days. Doubled as the foundation for new team standards around AI-augmented development.

The Product Play

Role
Technical Architect & Developer
The Mission
Eliminate latency and noise in sports injury tracking by building a real-time, automated intelligence platform.
The Build
Serverless full-stack solution using Supabase (Postgres), React + Vite, and Cloudflare Pages. Automated ingestion engine via Google Apps Script to normalize NFL API streams and trigger personalized alerts via Resend.
The Outcome
Live at fantasyinjurytracker.com, processing injury updates every few minutes with zero manual oversight.

The Speed Play

Role
Technical Product Manager
The Mission
Launch a complex compliance product for a major technology client under aggressive deadlines.
The Build
Delivered requirements, coordinated engineering sprints, and managed client expectations end-to-end.
The Outcome
Delivered in 1.5 months against an industry-standard 3 to 6 month timeline, unlocking new revenue streams and a multi-year client engagement.

The Scale Play

Role
Director of Operations
The Mission
Design and execute a statewide K-12 COVID testing program from scratch.
The Build
Built operational models, logistics workflows, and data pipelines serving thousands of students across multiple districts.
The Outcome
Scaled to $7.8M annual revenue; improved sample collection throughput by 40%; reduced invalid samples by 15%; expanded contracted school systems by 71%.
Tools & Technologies
AI-Augmented Engineering
Claude Code, Codex, Cursor, GitHub, OpenAI API, Groq API
AI Systems & Data Pipelines
LLM Orchestration, AI Ranking/Evaluation, Automated ETL, Structured Data Extraction
Full Stack & Cloud Architecture
Supabase, Cloudflare D1, Clerk, Netlify, Stripe API, JavaScript/React
Data Infrastructure & BI
Snowflake, Google Apps Script, Tableau, PowerBI, Looker Studio