Ben Pritchard

Edtech startup · Remote, UK · Role

Training platform for stock traders. Led full-stack development of a gamified trading simulation.

• Led full-stack development of a gamified trading simulation platform.
• Designed and built object-oriented back-end services in Java, including REST APIs and WebSocket integrations for real-time event handling.
• Built front-end interfaces with a focus on usability and clean design.

Nestwork · London, UK · Role

Pre-seed edtech startup helping students find and secure their first job.

• Built key infrastructure and the back end for the web platform, which grew to 8,000+ users within a year.
• Owned the AI career coach: integrated an LLM with the platform using custom functions, generating revenue from clients including Samsung, Loughborough University and B2C customers.
• Developed the onboarding workflow, contributing to a 15% reduction in drop-off through the conversion funnel.
• Worked cross-functionally with product, sales and marketing in an agile process, turning user feedback into rapid iterations.

Stride Ecosystem · London, UK · Role

Edtech startup helping entrepreneurs gain the skills they need to succeed.

• Led development of a subscription-based video platform, architecting a scalable, secure back end and a responsive front end, and managing a cross-functional team to a tight timeline.
• Designed, built and hosted the platform's infrastructure, managing and monitoring deployments from the Linux command line.
• Designed and deployed authentication and data protection measures to meet GDPR requirements.

Loughborough University · Loughborough, UK · Education

First Class Honours. Final-year project on using deep learning to predict and trade the stock market.

• Built an end-to-end pipeline in Python and TensorFlow: an LSTM retrained on a rolling window, fed by engineered technical indicators (RSI, MACD, ATR and others), driving an automated trading bot on the Alpaca API and containerised with Docker.
• Compared four LSTM architectures, with and without attention, and tuned hyperparameters with Keras Tuner; a single-layer LSTM without attention gave the lowest error and bias.
• Designed the risk layer: volatility-adjusted dynamic confidence thresholds, confidence-scaled position sizing, and ATR-based stop-loss and take-profit.
• Backtested on SPY, QQQ and TSLA: 10.45% CAGR with a 68% win rate on SPY (2022–2025) and 12.63% CAGR on QQQ, with weaker results on the more volatile TSLA.

Queen Mary's College · Basingstoke, UK · Education

Business, Computer Science and Mathematics.

• Business: A*
• Computer Science: A
• Mathematics: B