Rafael Evangelista

Dancing Blades Combat System

Unity, C#, VFX Graph, Post-Processing, State Machines, Spline/Procedural Motion

Overview
I built Dancing Blades as a short Unity prototype to explore how far I could push readable combat while controlling six floating blades at once. The demo is driven by a state machine + combo flow for movement and attacks, with each blade moving through a procedural, spline-based system that generates offsets around the player and can be tuned through parameters like speed, distance, and hover. I also leaned into stylised feedback, VFX Graph streak/ribbon effects and a post-processing pass, to sell dashes and fast transitions, then finished with audio and animation tweaks to make impacts feel more intentional.Skills:
- Designed and implemented the full combat + movement system architecture (state machine + combos).
- Implemented procedural blade movement and tuning controls.
- Created speed VFX and post-processing pass for movement/combat feedback.
- Integrated and polished audio + animations; assembled the demo build.Timeframe: Built over 1 week.

Voice-Controlled Reinforcement Learning Pet

Unity, C#, Reinforcement Learning, Q-Learning, Voice Commands, NLP, Human Feedback, JSON, AI Gameplay

Overview
I built this voice-controlled pet prototype in Unity because I wanted to explore what it would feel like if a companion could genuinely learn from the player in real time, instead of relying on fixed, scripted behaviours. The player gives commands through voice, the pet tries to interpret and carry them out, and training happens live using both automated learning and simple positive/negative feedback from the player while the action is happening. Over time, the pet can improve how it responds to commands, and the project includes saving/loading trained pets plus basic training graphs so I could track whether learning was actually improving.Skills:
- Implemented a Unity-based Q-learning system that learns from outcomes over time, designed for actions that take a duration to complete.
- Built a voice command pipeline in Unity and an intent-matching layer to map speech to actions.
- Integrated APIs to support dynamic command definitions and reward rules stored in JSON.
- Added human feedback controls (positive/negative reinforcement) during training to correct unwanted behaviour.
- Implemented save/load for trained models and logged telemetry to visualize rewards, success rate, and convergence.Timeframe: Oct 2025 – Dec 2025 (Class Project).

Bloom or Doom

Unity, C#, Mobile Development, Gyroscope Input, One-Handed Accessibility, Two-Player Multiplayer, NFC Integration, Rapid Prototyping, Team Collaboration

Overview
Bloom or Doom was created for a game jam during a Tencent-hosted jam at Kingston University. My teammates and I wanted to challenge ourselves by shipping an ambitious mobile concept under a tight deadline. We designed the game around single-handed accessibility, using gyroscope controls so players could play comfortably with one hand, and I worked on integrating some of the trickier tech pieces, NFC interaction and a two-player connection, so the core loop still felt stable and demo-ready. The jam environment pushed us to communicate fast, test constantly, and make quick scope decisions based on what actually worked in play.Skills:
- Built gyroscope-based controls to support single-handed mobile accessibility.
- Integrated NFC-based interaction as part of the gameplay concept.
- Implemented and tested a two-player connection feature to support multiplayer play.
- Collaborated in a jam team, iterating rapidly through short test cycles and feedback.Timeframe: Game jam (72 hours).