Passionate data scientist with a nuclear physics background, transitioned into building cutting-edge AI-driven trading systems and generative AI solutions. Developed a high-performance RL trading agent (EV >0.72), a BTC trend predictor (68% accuracy in the trend prediction), and over 500 PineScript strategies, alongside managed portfolio with 25-30% annual returns across last 3 years. Ready to deliver scalable, high-impact AI solutions for quant funds and innovative projects. 🚀
Star/fork these – sponsors get exclusive models & updates. Hire me on Upwork for custom versions!
RL Trading Agent for BTCUSDT (EV >0.72) – High-Performance Reinforcement Learning
Overview
Developed a sophisticated PPO-based reinforcement learning agent for BTC/USD trading on a 15-minute timeframe, leveraging a custom Gym environment with a proprietary reward function. Integrates 80+ high-signal entry points (TP(ATR)=SL(ATR), win rate >60% in backtests) derived from multi-timeframe market patterns (M15, H1, H4, D1).
Key Features
- Advanced Architecture: Utilizes a TCN feature extractor (64-256 channels, dilations [1,2,4,8], kernel=10) with action masking for trade compliance.
- Performance: Achieves robust profitability from epoch 1 (32% profit at 10% drawdown in bullish phases).
- Scalability: Optimized for any crypto exchange or MT5 broker without slippage for large enough portfolio.
Tech Stack
- Python, PyTorch, Stable-Baselines3
- Custom TCN, Pandas, NumPy, TA-Lib
- Bybit/MT5 APIs for real-time execution
Links
- Repo: RL BTC Trading Agent
- Monetization: Contact me to customize this agent for your fund’s trading strategy.
BTC Trend Predictor (68% Accuracy, 0.67 F-Score) – Real-Time Multi-Timeframe Forecasts
Overview
Built a hybrid TCN-LSTM model with MultiHead Attention for predicting the next candle’s direction across 4h, 1d, and 1w timeframes. Achieves 67-68% accuracy and 0.67 F-score on out-of-sample data, surpassing random baselines by 16-18%, with streaks of 95+ correct predictions.
Key Features
- Robust Predictions: Processes high-frequency OHLCV data via CCXT, computing 20 technical indicators (RSI, MACD, Bollinger Bands).
- Data Pipeline: Applies log1p and tanh normalization with rigorous NaN/gap detection and forward-filling for stability.
- Deployment: Runs 24/7, delivering consistent real-time forecasts for multi-timeframe trading.
Tech Stack
- Python, TensorFlow, Keras-TCN
- Pandas, TA-Lib, CCXT
Links
- Repo: Crypto BTC Trend Prediction (currently private)
- Engage: Follow my X profile for live prediction updates or sponsor for access to the full dataset.
PineScript Strategies (500+ Scripts) – Multi-Timeframe TradingView Systems
Overview
Crafted 500+ PineScript v6 strategies for TradingView, focusing on multi-timeframe analysis (M15, H1, H4, D1) and market phase decomposition. Includes 80+ core strategies with win rates of 55-75% and sharpe ratio of 1.3-1.8.
Key Features
- Robust Design: Utilizes RSI, Bollinger Bands, EMA, HMA, and ATR without curve-fitting for market adaptability.
- Multi-Timeframe: Combines M15 signals with H1/H4/D1 context for high-signal entries.
- Customization: Ready to enhance any trading system with tailored, high-performance strategies.
Tech Stack
- PineScript v6
Links
- Repo: PineScript Strategies
- Monetization: Hire me on Upwork to develop or optimize your trading strategies.
Trading Bots Portfolio (50+ Scripts) – Crypto Futures
Overview
Designed and deployed 50+ trading bots for BTCUSDT futures, achieving 20-25% annual returns with 10% drawdown while managing a $330K portfolio. Focused on diverse, rigorously backtested trading strategies.
Key Features
- Performance: Delivered consistent returns through automated systems optimized for crypto futures.
- Backtesting: Employed rigorous testing to ensure strategy robustness across market conditions.
- Experience: Demonstrates deep expertise in building and managing high-stakes trading systems.
Tech Stack
- TSLab, Backtrader
Details
- Showcases my experience in developing and managing automated trading systems for crypto futures.
A comprehensive set of technical and interpersonal skills honed through advanced AI-driven trading systems, data engineering, and generative AI projects, underpinned by a nuclear physics background that drives analytical precision and problem-solving.
| Category | Tools & Expertise |
|---|---|
| AI/ML | - Frameworks: Proficient in PyTorch, TensorFlow, Keras and Stable-Baselines3 for building and optimizing neural networks and reinforcement learning models. - Architectures: Expertise in designing hybrid TCN-LSTM-CNN models with MultiHead Attention and custom feature extractors for time-series prediction. - Reinforcement Learning: Advanced PPO implementation with action masking, custom reward shaping, and gradient hooks for stability. - Model Optimization: Skilled in dataset preparation, hyperparameter tuning, layer normalization, and optimization to mitigate overfitting and ensure robust performance. - Predictive Modeling: Developed models achieving high accuracy and F-scores for financial forecasting. |
| Trading Systems | - Strategy Development: Expert in crafting robust trading strategies using python, PineScript or TSLAB, with multi-timeframe analysis (M15, H1, H4, D1) and market phase decomposition. - Automation: Proficient in Pinescript, TSLab and Backtrader for building automated trading bots, with rigorous backtesting for performance validation. - API Integration: Seamless integration with any crypto exchange and MT5 APIs for low-latency execution and real-time market data processing. - Risk Management: Skilled in designing systems with dynamic TP/SL and low drawdown strategies for high-stakes environments. - Market Analysis: Deep understanding of market phases, technical indicators and multi-timeframe analysis. |
| Data Engineering | - Data Processing: Advanced proficiency in Pandas, NumPy and Matplotlib for handling high-frequency financial datasets. - Feature Engineering: Expertise in computing technical indicators using TA-Lib and implementing robust normalization. - Time-Series: Skilled in multi-timeframe data preprocessing, imputation, and quality checks for stable model inputs. - Data Pipelines: Built scalable pipelines for real-time OHLCV data fetching and processing. - Performance Optimization: Optimized data workflows for high-throughput environments, achieving high FPS on GPUs across parallel environments. |
| Generative AI | - Leadership: CTO at FlirtAGPT, driving the development of innovative AI products for conversational and creative applications. - Model Design: Experienced in building and deploying generative AI models, leveraging deep learning frameworks for scalable solutions. - Product Development: Skilled in end-to-end AI product lifecycle, from ideation to production deployment, with a focus on user-centric outcomes. - Innovation: Applied cross-disciplinary insights from trading and physics to create novel generative AI use cases. |
| Soft Skills | - Analytical Thinking: Leveraged nuclear physics background to tackle complex problems with precision and rigor. - Problem-Solving: Developed innovative solutions for trading and AI challenges, balancing performance and scalability. - Communication: Effectively translate technical concepts to stakeholders, as demonstrated in client engagements on Upwork and X posts. - Adaptability: Successfully transitioned from nuclear physics to data science, mastering diverse tech stacks and domains. - Collaboration: Open to partnerships with quant funds and startups, with a track record of delivering tailored solutions. |
My journey from nuclear physics to data science, building innovative AI-driven trading systems and generative AI solutions:
- 2021: Developed first predictive model for gold trend forecasting and built initial forex trading bot, laying the foundation for automated trading systems.
- 2022–2023: Created and managed a TSLab portfolio with 50+ unique algorithms, focusing on crypto futures and rigorous backtesting for performance.
- 2023–2024: Designed hundreds of PineScript v6 strategies for TradingView and advanced predictive models, emphasizing multi-timeframe analysis and robust forecasting.
- 2025: Engineered a high-performance RL trading agent and a state-of-the-art predictive model, deployed in real-time for BTCUSDT trading.
I’m passionate about building cutting-edge AI trading systems and generative AI solutions. Let’s create value together!
- Quant Funds & Startups: Partner with me to develop scalable, high-impact trading algorithms and AI models. DM on LinkedIn or X.
- Custom Solutions: Hire me on Upwork for tailored RL agents, predictive models, or trading strategies.
- Sponsorship: Gain early access to exclusive models and datasets – Sponsor Now.
Star my repos to stay updated on the latest AI trading innovations! 🚀
