🧰 Also working with

_{C++  ·  Rust  ·  OCaml  ·  R  ·  SQL  ·  Shell}

Track Record 📈 Five years of my own broker-verified trading: ₩65.9M → ₩241.9M of total value ($49.3K → $181.1K) · 3.67× MOIC — compounding at 34.3% a year time-weighted (32.1% IRR) versus 15.1% for the KOSPI and 14.8% for the S&P 500, a +19.2 pp annual edge and +22.4% Jensen's alpha. ₩176.0M total profit with an 87% hit rate, 12.6× profit factor, Sharpe 0.94 / Sortino 2.20 and only 12 bp all-in trading cost; best trade +337% (HD Hyundai Heavy Industries). Every won reconciles to the Kiwoom statements, published in Korean and English. _{🧰 Python · pdfplumber · pikepdf · openpyxl · TWR / Modified Dietz · XIRR · Sharpe / Sortino / Jensen α · KOSPI · S&P 500 · Nasdaq-100 benchmarks}

Automated-DCF-ML
Engineers SEC Company Facts / XBRL data and fits Ridge & Elastic Net forecasts for six value drivers (revenue growth, EBIT margin, tax, D&A, CapEx, NWC intensity); converts them to NOPAT and FCFF, then a DCF to enterprise / equity value and fair value per share. Chronological training (through 2018) and testing (2022+).
_{🧰 Python · scikit-learn · SciPy}

InsightStream ⭐
DigitalOcean Gradient™ AI Hackathon — AI observability platform that streams logs through Kafka and uses Llama 3 to detect anomalies and summarize incidents in real time.
_{🧰 Kotlin · Spring Boot · Kafka · Next.js}

Revenue Dip Lab
Backtests buying companies with resilient revenue after sharp price drawdowns — five-panel research workspace, 3D CAGR surface sweeps and out-of-sample factor search on SEC EDGAR + Yahoo Finance data.
_{🧰 Python · FastAPI · JavaScript · Web Workers · SQLite}

Signals to Alpha
Research program turning real-economy signals into testable alpha factors — a U.S. K-Food alternative-data factor model with point-in-time signals, placebo tests and bias audits.
_{🧰 Python · statsmodels · SciPy · Jupyter}

MIDAS Terminal
Automated equity research terminal — ticker → market data → SEC EDGAR filings → financial analysis → peer comparison → rule-based catalysts & risks, exported as English / Korean Word research reports; async fetching with single-flight TTL caching and rate limiting.
_{In production use at MIDAS Capital
🧰 Python · FastAPI · JavaScript · Plotly · SQLite · SEC EDGAR}

Nanny Mira — Child Temperament AI
Parent-reported child temperament web app in 4 languages (KO / EN / RU / PT-BR): an age-adapted 28-question assessment, 40 expandable analysis cards and DeepSeek LLM parenting guidance grounded with PubMed references; Prodamus card checkout.
_{Represented by: Илона Масельскене, Russia partner
🧰 Node.js · DeepSeek API · Low-Cost LLM Inference · RAG · Prompt Optimization · Context Caching · Token Budgeting · Usage Quotas}

Superdupercatai
AI comics & animation channel — AI-generated content production, growth and monetization analytics (70K followers, 20M peak views, ~1M average views / video).
_{🧰 Generative AI · content analytics}

Extriver Partners
Product & business modeling — creator audience geography, engagement and sector demand to evaluate vertical-software concepts; subscription, conversion and revenue-share scenarios; internal partner pipeline and research-to-execution workspace. Distribution tested through an owned short-form account (70K followers, 20M peak views, ~1M average views / video).
_{Partners: LoveSoo, TikTok · YouTube · Linktree  ·  Stick Stories  ·  Илона Масельскене  ·  Superdupercatai  ·  틱톡★꽐루  ·  Bushra  ·  BTS BORA WORLD (보라TV), TikTok  ·  Tushar Shetty, YouTube · Facebook
🧰 Node.js · Socket.io · Cloudflare R2}

🔬 Modeling libraries

📐 ML models

🤖 LLMs & AI engineering

📈 Financial data

⚙️ MLOps & data

🧩 Application stack

• 📐 Regularized regression — Ridge / Lasso / Elastic Net with CV, Pipeline + StandardScaler
• ⏱️ Leak-free validation — Chronological training and out-of-period testing to avoid look-ahead bias
• 📊 Statistical testing — OLS with statsmodels, Spearman / Pearson correlation, curve fitting with scipy.optimize
• 🧮 Factor models — Alternative-data signals as equity factors; multi-factor strategy fitting and backtesting
• 🛰️ LLM applications — RAG pipelines with vector search, LoRA / PEFT fine-tuning, real-time anomaly detection and research-grounded guidance
• 💹 Finance — DCF / FCFF · public comps · 3-statement & LBO · M&A diligence · EV-to-equity bridges