Prize Categories Entered

• 

  Best Use of Gemini API

  Submitted

  We used Google Gemini API as the core intelligence behind Hermeneus, our AI-powered human-to-robot interpreter. Gemini combines natural-language instructions, visual observations, and robot capability information to understand human intent, reason about the environment, and generate structured JSON action plans. Our custom Python interpreter then validates and translates these plans into robotic commands, spatial coordinates, movement trajectories, and verbal responses, with Unitree humanoid robots as our target platform. The key innovation is adaptability: Rather than relying on predefined commands, Gemini enables Hermeneus to interpret different human requests and generate context-aware robotic actions based on the surrounding environment.

• 

  Best Use of Tiger Data

  Submitted

  We used Tiger Data (TimescaleDB/PostgreSQL) as the persistent memory and execution history for Hermeneus, enabling us to track how human instructions are interpreted and translated into robotic actions. For each interaction, Tiger Data stores timestamped inputs, environmental observations, Gemini's reasoning outputs, structured action plans, generated trajectories, and interpreter results. By combining PostgreSQL with time-series data capabilities, we can reconstruct past interactions, analyze how the robot's decisions evolve as its environment changes, and maintain a traceable history of AI-generated actions.

• Event prize

  Best Overall

  Submitted

  Hermeneus bridges the gap between human communication and robotic execution. Using Google Gemini, our system interprets natural-language instructions and visual observations, translating them into structured robotic commands, movement trajectories, and verbal responses. Built with Python, Sentry, and Tiger Data, Hermeneus combines multimodal AI, deterministic interpretation, and persistent execution history into a robot-independent architecture. Rather than building another chatbot or preprogrammed robot, we're developing an adaptive intelligence-to-action layer that makes advanced robotics accessible through natural human interaction.

• Event prize

  Sentry Raffle

  Submitted

  Hermeneus uses Sentry to monitor the entire AI-to-robot interpretation pipeline, tracking Gemini's reasoning, generated action plans, interpreter execution, and potential errors. By tracing each stage, Sentry helps us identify where instructions are misunderstood, commands fail validation, or trajectories cannot be generated. This visibility is especially important in physical AI, where incorrect interpretations can lead to unintended robotic actions. Sentry makes Hermeneus more transparent, debuggable, and reliable as we work toward real-world robotic execution.