Financial engineering, taught on real market data.

Bonton Academy teaches engineers to build trading and investment systems that hold up in real markets, on the same infrastructure Bonton AI runs itself. The lessons are open. The first cohort is in preparation.

Who it is for

• Researcher. Data scientists and ML engineers moving into quantitative research. How market data is produced and distorted, how to measure return and risk without fooling yourself, portfolio construction under real constraints and an evaluation that does not overstate a result.
• Developer. Engineers building trading and investment software. Market data pipelines, limit order books, execution costs and a backtest whose fills match what the market would have given you.
• Executive. Heads of risk, investment and technology who decide from results. What a backtest can and cannot tell you, how to read a risk report and the question to ask before you sign off a model or a vendor. No code.

Open lessons

• Why 25-sigma days keep happening. The five things every return series does, the model that captures them and what they do to the risk number a bank reports every day.
• The optimizer is an error maximizer. Markowitz, eleven portfolios and eighteen years out of sample: what estimation error does to an optimizer, what shrinkage, constraints and clustering fix, and what they cannot.

How it is taught

• Real data, including real defects. Every topic is worked on real prices: equities, ETFs, FX, commodities, crypto, tick data and order book snapshots. Learning to distrust a data feed is part of the work.
• Derived, not recited. The math is written out step by step, next to the numbers it produces. A method you did not derive is one you cannot repair when it stops working.
• Honest results. Lessons show what fails as well as what works, with the evaluation that tells the two apart. A negative result, competently established, is a result.
• Judged on the reasoning. Work is judged on three things: the code, the number it produces and two or three sentences on what you concluded and what would make you doubt it. The third carries the most weight.
• No strategies to copy. We teach how to build and judge a strategy. We do not hand out strategies to run and we never promise returns.

Who teaches it

Bonton Academy is created by Robert Yenokyan, founder of Bonton AI. He builds the research and trading infrastructure Bonton AI runs, and the lessons teach what that work uses.

He co-authored two textbooks, Ordinary Differential Equations in Exercises and Problems, Parts 1 and 2. He taught Data Visualization at Yerevan State University for five years and now teaches Financial Engineering there.

The Bonton Academy course is much broader and much deeper than a university course. It is also how Bonton AI trains its newcomers and its own team.

What the course covers

1. Markets, instruments and how market data is made
2. Data engineering: ticks, bars, order books and the defects inside the data
3. Returns, risk and performance measurement
4. Stylized facts, volatility models and tail risk
5. Portfolio construction under real constraints
6. Backtesting: walk-forward tests, leakage and the cost of trying many ideas
7. Execution and trading costs: slippage, queue position, fees and fills
8. Forecasting: classical models, machine learning and foundation models
9. Labels, features and leakage in financial machine learning
10. Mean reversion, momentum and factor research
11. Market microstructure and order book analytics
12. Hedging: FX, commodities, scenarios and stress tests
13. Regimes and scenario risk
14. Research design and a project built and defended on real data

Next

A founding cohort is in preparation. Join the waitlist or write to info@bonton.ai.

Bonton Academy is the education arm of Bonton AI, a quantitative research and development firm in Yerevan, Armenia, founded by Robert Yenokyan.