Shigemi Yagi

MQL5 / MQL4 developer with 7+ years of hands-on experience building Expert
Advisors, custom indicators, and machine-learning-driven trading systems
for MetaTrader 4 and MetaTrader 5. I also work daily as a data scientist,
which means the EAs and indicators I deliver are backed by the same
statistical rigor I apply to my day job — proper out-of-sample validation,
walk-forward testing, and overfitting checks, not just curve-fitted
backtests.

================================================
Core Expertise
================================================
- MQL5 / MQL4 — Expert Advisors, custom indicators, scripts, libraries
- Strategy Tester — single-pass and genetic optimization, walk-forward,
multi-symbol setups
- ONNX integration in MT5 — native inference of ML models (LightGBM,
RandomForest, MLP) inside Expert Advisors
- Python — model training and feature engineering with pandas, numpy,
scikit-learn, PyTorch, LightGBM
- Pine Script v5 / v6 — TradingView indicator and strategy development
- Risk management modules — prop-firm-style drawdown controls
(FTMO-compatible daily loss / total loss / news filter)
- Multi-timeframe and multi-symbol architecture, ensemble logic,
asymmetric ATR-based TP / SL

================================================
Selected Project Examples
================================================
- 5-logic ensemble Expert Advisor for MT5 with 3-timeframe majority voting,
ATR-based asymmetric take-profit / stop-loss, dynamic lot sizing, and a
full prop-firm-compatible risk control layer (daily loss, total loss,
trading-hour filter)
- Multi-pair AI Expert Advisor running LightGBM models exported to ONNX
and inferred natively inside MT5, simultaneously trading USDJPY,
EURUSD, and GBPUSD with a confidence-threshold filter
- Multi-timeframe "Confluence" Expert Advisor combining a regime-detection
model with three independent signal models, each trained separately and
combined by a voting layer
- Pine Script library with a usable RandomForest implementation, allowing
TradingView users to run tree-based models directly from chart code

================================================
What I Can Build for You
================================================
- Custom Expert Advisors from your written rules or existing strategy
- Conversion of TradingView / Pine Script strategies into MT5 EAs (or the
reverse direction)
- Machine-learning-driven EAs: I train the model in Python, export to ONNX,
and integrate it natively into MT5 — no external bridge required
- Custom indicators (single or multi-timeframe), alert systems, and
dashboard panels
- Strategy backtesting, walk-forward analysis, and parameter optimization
- Prop-firm-ready risk management modules (FTMO, MFF, FundedNext-style
rule sets)
- Code review, debugging, and refactoring of existing MQL4 / MQL5 code

================================================
Why Work With Me
================================================
- Three-stack capability — MQL5, Pine Script, and Python ML in one
developer. Cross-platform projects (TradingView indicator + MT5 EA, or
ML model + EA integration) do not need to be split across two people.
- Data-science background — I work as a data scientist outside of MQL5
freelancing. I treat backtests with the same skepticism I treat any
model: out-of-sample, walk-forward, and stress tests come standard.
- ONNX in MT5 — native ML inference inside MetaTrader 5 is still rare on
the platform. I have shipped working implementations.
- Direct English communication — TOEIC 960. No translation tools, no
back-and-forth misunderstandings on requirements.

================================================
How I Work
================================================
- I read your specification carefully and ask clarifying questions before
writing a single line of code. Most disputes come from ambiguous
requirements, and I try to resolve them up front.
- Milestones for larger projects so you can review progress before final
payment.
- Clean, commented MQL5 code that you (or another developer) can maintain
later. I do not deliver black-box binaries unless you specifically
request a protected version.
- Reasonable post-delivery support window for bug fixes related to the
original specification.

================================================
Get in Touch
================================================
Feel free to send a Job Request with your project details — preferably
including the rule logic, target symbol(s), timeframe, and any reference
indicators or screenshots. I will reply with a realistic time estimate
and any clarifying questions before we agree on terms.

Grafici & Idee