Projects
Hi Buddy – Mental Health Buddy
An experimental multimodal companion focused on emotional state understanding and supportive conversation.
Multimodal Research Assistant
A prototype assistant exploring multimodal foundation models for navigating complex research literature.
Fun with Chess
A game concept for kids and beginners, using story and gentle rules to make learning chess less intimidating.
Ascent Everest
A 3D journey from basecamp to summit, exploring how game mechanics can support focus, planning and resilience.
Hi Buddy – Mental Health Buddy
Hi Buddy is an experimental multimodal companion that tries to understand a person’s emotional state from text, speech, facial expression and video, and then responds with supportive, bounded conversation.
The project is a playground for thinking about wellbeing, guardrails and how AI agents should behave when they notice someone is struggling. It focuses on safe defaults, escalation boundaries and transparent limitations rather than open‑ended “therapy”.
Multimodal Research Assistant
This prototype explores how multimodal foundation models can help researchers reason across text, figures and diagrams in complex technical domains. The goal is to reduce the friction of moving from papers to practical insights and experiments.
We experiment with workflows that combine document understanding, citation tracing and interactive querying, inspired by recent work on vision foundation models and multimodal agents that can see, think and act.
Recent Advances in Vision Foundation Models CVPR 2025:
- Advancing Multimodal LLMs: From Seeing to Understanding and Acting – Zhe Gan
- Multimodal Reasoning for Visual‑Centric Long‑Horizon Tasks – Zhengyuan Yang
- See. Think. Act. Training Multimodal Agents with Reinforcement Learning – Linjie Li
- Towards Multimodal AI Agent That Can See, Think and Act – Jianwei Yang
Fun with Chess
Fun with Chess is a game concept for kids and new players, designed to make learning chess feel less intimidating and more playful. It experiments with simplified rules and visual feedback to support beginners without overwhelming them.
Ascent Everest
Ascent Everest is a 3D game concept based on a journey from basecamp to the summit, where players need to plan, adapt and overcome small challenges on the way.
It explores how game mechanics and AI‑driven events can help players practise focus, decision‑making and perseverance in a playful setting.
About us
inventVALLEY is a lean team combining applied AI research, product engineering and a curiosity‑driven approach to learning and play.
Our work is led by an applied AI researcher and engineering leader with experience delivering large‑scale conversational AI, speech recognition and recommendation systems in industry, with a focus on multilingual models, AI evaluation infrastructure and responsible AI.
Contact
If you are interested in collaborating, exploring VeriTuring, or discussing applied AI projects, we’d be happy to talk.
Email: contact@inventvalley.com
Organisation: inventVALLEY Limited, London, United Kingdom.