Machine Learning & AI
Deep reinforcement learning, neural networks (ANN/CNN), LSTM/RNN and ensemble models — plus production LLM agents. From portfolio RL and volatility models to agents that reason over live trading data.
Data Scientist at eToro. I build production AI across the stack — deep reinforcement learning, neural networks and LSTM models, LLM agents, and on-chain DeFi systems — on a foundation of real trading, hedging and execution expertise. Then I sit with the desk and get them to adopt it.
I didn't jump to AI. I built the quant and ML foundations, learned the revenue side, and now ship AI and DeFi on top of both — which is exactly what makes the work credible.
Built hedging algorithms with Random Forest and LSTM models, ran execution research, and modelled volatility and regime change with GARCH on full L2 order-book data.
Shipped an A/B-tested neural-network churn model and deep-RL portfolio construction, plus revenue and client-engagement analytics that connect models to the P&L.
Ship production AI agents and on-chain DeFi systems for trading — and embed with teams to get the whole desk adopting AI safely.
Advanced ML. Real trading and hedging. Modern DeFi. Most people have one — the value is in holding all three, and in getting teams to actually adopt what I build.
Deep reinforcement learning, neural networks (ANN/CNN), LSTM/RNN and ensemble models — plus production LLM agents. From portfolio RL and volatility models to agents that reason over live trading data.
Real quant-finance depth: hedging and execution algorithms, derivatives pricing, market microstructure, GARCH/regime models and multi-billion-dollar reconciliation. The domain knowledge that makes the AI trustworthy.
Tokenizing assets on-chain via mint and burn, writing Solidity smart contracts and on-chain liquidation bots, and engineering them to resist MEV and sandwich attacks across EVM and Solana. Modern crypto, built robustly.
I embed with teams doing manual work, build the agent around their workflow, automate it — then teach them to run and extend it themselves. Build, hand over, multiply.
Production work across ML, AI agents, DeFi and a full learning platform. Internal systems are described at a deliberately high level — no confidential figures or names.
A body of production machine-learning work applying deep learning, reinforcement learning and classical ML to real trading and client problems.
An AI-powered platform that explains multi-billion-dollar daily position breaks across multiple liquidity providers in seconds — built with the team that used to do it by hand.
An automated agent that reconciles and reports hedging cost across every asset-class book each day, with anomaly detection and self-service dashboards — now run by the desk itself.
Hands-on DeFi engineering: tokenizing assets on-chain and building the smart-contract infrastructure around them — shipped and working.
A complete learning platform that takes someone from their first order through to mathematically rigorous advanced material — 21 live courses across trading, derivatives, execution, microstructure, portfolio construction, crypto/DeFi and compliance.
In 2026, the scarce skill isn't writing the agent — it's getting a non-technical desk to trust it, use it, and own it. My agents don't sit on my laptop: I embed with the team, build around their real workflow, automate the manual grind, and then teach them to run and extend the tool themselves — so my work multiplies instead of creating a dependency on me.
Watch the manual work, learn the edge cases, map how they actually operate.
An agent that fits their workflow — explainable, with guardrails they can trust.
The slow, error-prone manual steps disappear; answers arrive in seconds.
Teach them to use, validate and extend it with AI — they own and update it themselves.
Teach the desk to ask reliable questions of internal AI agents — and understand the limits of each answer.
Access docs, follow setup steps, and report issues — without needing to be a developer.
Practical, safe setup for power users to edit scripts and debug with AI in the loop.
Query and summarize trading data responsibly, with the guardrails that keep it safe.
How to avoid over-trusting AI — verification habits, and a clear "what not to do" for sensitive data.
Ongoing, low-friction support that meets people in their real workflow — reducing fear, building momentum.
Advanced ML, production AI agents, DeFi engineering — or getting a team to actually adopt AI. Happy to compare notes.