🎓 Honours B.Sc. Computer Science & Mathematics (Statistics Minor) @ University of Toronto
📍 Toronto, Canada
💡 Interests: Machine Learning, Data Engineering, Risk & Finance, Systems

I’m a CS & Math student with a strong focus on machine learning, data systems, and risk-aware modeling. I enjoy building production-style pipelines end-to-end — from raw data ingestion and feature engineering to evaluation, deployment artifacts, and documentation.

My work spans reinforcement learning for trading, credit risk modeling, large-scale time-series forecasting, and software/code security auditing. I care deeply about reproducibility, clean architecture, and turning complex models into systems that can actually be used.

Balancing Profit and Risk: Hybrid RL for Algorithmic Trading
Apr 2025 – Sep 2025

• Built a modular RL trading framework with risk-aware reward functions (volatility, drawdown, Sharpe-hybrid).
• Implemented Q-Learning and DQN agents with stabilizing techniques.
• Achieved +27% return improvement, Sharpe ≈ 1.93, and reduced max drawdown from 30% → 15% across 53 equities/ETFs.

Python · FastAPI · LangChain · Backend Systems

• Built TechnationAI, an AI assistant project focused on clean APIs and modular backend design.
• Designed components for tool usage, prompt orchestration, and extensibility for future features.

🔗 https://github.com/BenjaminADecosta/TechnationAI

Python · PySpark · AWS EMR/S3 · scikit-learn

• Built a production-style ETL pipeline using PySpark on AWS EMR to transform raw credit data into curated Parquet feature datasets stored in S3.
• Designed a config-driven architecture (YAML) separating data contracts, pipelines, training, and artifacts for reproducible runs.
• Implemented feature engineering + dataset standardization to support scalable training workflows.

🔗 https://github.com/BenjaminADecosta/AWS-Credit-Risk-Pipeline

iOS · Accessibility · 3D Printing · Product Design

• Accessibility-focused app empowering users (especially those with disabilities) to create custom 3D-printable assistive devices.
• Built around user-driven customization with a focus on inclusive, practical everyday tooling.

🔗 https://github.com/jacobamobin/Ability

Pandas · SARIMA · Ridge Regression · scikit-learn

• Forecasted weekly NYC EMS demand using 1.2M+ incident records and weather data.
• Implemented automated backtesting and 26-week forecasts with borough-level confidence intervals.

🔗 https://github.com/BenjaminADecosta/EMS-Incident-Forecaster

Security Code Auditor (Contract) — Hssndusqooq LLC

• Audited a React/Firebase creator platform (auth, storage rules, APIs, logging).
• Identified security & privacy gaps with file-level evidence and repro steps.
• Delivered an executive-ready remediation roadmap prioritizing real-world risk.

Machine Learning Developer — QAIFS @ UofT

• Built reusable regression, classification, and forecasting modules.
• Developed modular Python data pipelines for cleaning, validation, and transformation.
• Collaborated via Git with reproducible workflows and experiment tracking.

Languages: Python, Java, SQL, JavaScript, C/C++, R
ML & Data: PySpark, Spark SQL, Pandas, NumPy, scikit-learn, PyTorch, Hadoop, Parquet
Cloud & Tools: AWS (EMR, S3, SageMaker), Git, Jupyter, VS Code, PyCharm, IntelliJ

• GitHub: https://github.com/BenjaminADecosta
• LinkedIn: https://linkedin.com/in/ben-decosta
• Email: benjamin.adam.decosta@gmail.com

⭐ Pinned repositories below highlight selected work in ML, data engineering, and applied systems.