Palo Alto–based startup Traini has raised $7.5 million to build what it calls pet “emotional intelligence,” backing a future where dog owners don’t just track steps and sleep, but have real-time, voice-like conversations with their pets through AI. The round will fund R&D, product iteration, and global rollout of its Cognitive Smart Collar, pitched as the world’s first generative AI dog “language” translator in wearable form.
Funding, backers, and mission
The round is led by Banyan Tree, Silver Capital, ZhaoTai Group, and NYX Ventures, with participation from Starting Gate Fund, Jade Capital, and tech investors including NVIDIA VPs, Julian Qian of Anthropic, Weihe Zheng of Cowin Capital, Jia He of Nanshan Capital, Peter Xu of Plug and Play China, and Zach Zhang of Edgewater Investments. Existing investors, including the Tao Foundation and Xiaomi cofounder Feng Hong, also joined, building on earlier backing from leaders at Google, Meta, and Palo Alto Networks.
Traini founder Arvin Sun stated, “Our mission is to develop an intelligence that reignites the natural instincts of our furry companions and builds a true spiritual bond between pets and their humans.”
Cognitive Smart Collar and PEBI platform
Fresh capital supports the launch of Traini’s first hardware product, the Cognitive Smart Collar, described as the world’s first cognitive pet wearable and the first human–dog “language” translation device built on generative AI, now available for pre-order via the Traini app and site. The collar runs on Traini’s core multimodal engine, PEBI (Pet Empathic Behavior Interface), which processes text, images, video, and audio.
By analyzing a dog’s vocalizations, facial expressions, and body language, PEBI turns those signals into human-style conversational output designed to feel like a two-way exchange. Traini says its models currently cover nearly 120 dog breeds with emotion translation accuracy of up to 94%, enabling applications from behaviour translation and emotion recognition to personalised services and pet-assisted healthcare.
Data, models, and emotional mapping
Traini’s system combines a Valence–Arousal model, a 3D Pet Emotion Model, and an Instant Emotion Vector to locate each dog’s emotional state in real time. These models are built on insights from more than 900 peer-reviewed animal behavior studies and trained on behavioral data from over 2 million dogs.
To refine emotional mapping, Traini uses a “human–pet vocal spectrogram comparison” method, comparing spectrograms of human speech expressing specific emotions with canine vocal patterns to improve matching. The collar fuses sound analysis with vital signals such as heart rate, temperature, and body movement to detect states like joy, anxiety, excitement, and distress, then translates them into “what the dog wants to say” in plain language and actionable owner feedback.
Scale, community, and retail reach
On the user side, Traini reports it has already served over 2 million dogs and generated more than 70 million views on YouTube. In 2023, it launched PetGPT, a natural language and behavior analysis model with 99 percent coverage across users, which lifted service engagement by 70%.
Earlier in 2025, Traini unveiled T-Agent, an AI-powered recommendation and purchasing system that autonomously identifies dogs’ real needs, effectively letting pets “make decisions” with owners as co-pilots. The company now works with nearly 40,000 local pet stores across the U.S. and has struck partnerships with leading smartphone and electric vehicle brands so owners can eventually interact with their dogs via mobile operating systems and in-car infotainment screens.
A “train-as-you-use” mechanism sends anonymised user–pet interactions back into its PPI (Pet Perception and Interaction) model, steadily expanding and refining the dataset. Investors describe Traini as a breakthrough example of vertical AI where deep data, multimodal models, and emotional use cases intersect.
