36Kr Exclusive | HSG China and Monolith Lead Investment in AI Health Hardware Company "Odyss" with Nearly 200 Million Yuan

Author | Zhang Ziyi

Editor | Yuan Silai

Yingke learned that the AI health hardware company "OdyssLife" (hereinafter referred to as "Odyss") has recently completed multiple rounds of financing in succession, with a total amount of nearly 200 million RMB. This round of financing was led by Sequoia China and Monolith respectively, and old shareholders Linear Capital and Creekstone continued to increase their investment.

The funds from this round of financing will be mainly used for product software and hardware R & D, global marketing, and team expansion.

At a time when the AI hardware track is highly sought after by capital, Odyss has entered a high - frequency scenario that has not been fully digitized - diet and health monitoring.

Odyss' first product, the Odyss N1, is the world's first Always - On smart necklace, integrating multi - modal sensing capabilities such as image, audio, and motion. It can perceive and record users' diet and exercise behaviors around the clock.

Pan Yuyang, the founder and CEO of Odyss, believes that in the daily high - frequency scenario of diet, existing smart wearable devices have physical limitations. Smartbands and rings have no vision and cannot perceive food; while CGM belongs to medical devices with a strong medical attribute and is naturally far from ordinary consumers.

In comparison, the advantage of the necklace lies in its light weight. A weight of just a few dozen grams is enough to achieve all - day "senseless wearing". It also has a natural wide - angle "god's eye view" and can be equipped with a low - power visual sensing camera module, solving the pain point that traditional wearable devices cannot accurately capture "diet data".

Specifically, the Odyss N1 has built a three - modal sensing system with "vision as the main, supplemented by audio and motion sensing". The core visual module abandons high - energy - consuming continuous video shooting and instead uses low - power "frame - taking" technology to accurately capture data from all scenarios, from dining in restaurants to home cooking, and can identify fine - grained information such as food types, volumes, and cooking methods.

At the same time, the audio and motion modalities are responsible for capturing semantic keywords in ordering and monitoring the user's metabolic state respectively, assisting in improving data accuracy. Based on this, the system will conduct in - depth learning on complex data such as calories, nutrients, and glycemic index, and finally generate personalized health intervention strategies for users based on their real - life states.

While many AI hardware startups reuse mature supply - chain solutions to reduce costs, Odyss, which has obtained sufficient "ammunition", chooses to continuously optimize its products by customizing components.

At the hardware level, in order to achieve low - power, high - precision visual sensing in a very small necklace volume, the company has customized a dedicated visual module and power supply system, and adopted a more costly titanium alloy CNC processing technology to further enhance the appearance texture.

At the software level, Pan Yuyang believes that the accuracy of general models in vertical tasks such as identifying food types and estimating calories has not reached the commercial standard.

Therefore, Odyss has built a proprietary link of "small - model pre - training + large - model post - training". The large model is responsible for identifying food types and cooking methods, which is a crucial "backup" guarantee for diet monitoring, and can retrieve the composition data of food from the database for calculation; the traditional computer vision (CV) algorithm has high accuracy and is specifically responsible for measuring food sizes and scanning codes.

Yingke learned that according to the calculations of the Odyss team, the Odyss N1 can achieve an accuracy rate of over 90% in identifying the calorie content of regular Western food.

In terms of business model, Odyss plans to adopt a combination of "hardware + subscription". In the early stage, the company plans to maintain a high - end positioning of "tech trendy products" on the hardware side, but reduce the initial software subscription fees, aiming to quickly expand the user base and establish category awareness by lowering the usage threshold.

In terms of target customer groups, Odyss plans to initially target high - net - worth groups such as "bio - hackers", technology and finance practitioners, and fitness enthusiasts as its core users.

Pan Yuyang believes that these seed users not only have a high sensitivity to data and strong payment ability, but more importantly, they are also the definers of "social currency".

"In the consumption logic of high - net - worth individuals, wearable devices often serve as identity tags, and no one wants to wear a copycat that ranks second in the market around their chest," Pan Yuyang said.

Odyss has formulated a high - frequency SKU iteration route in its product planning. The company plans to launch a new series every 3 - 4 months. Through high - specification CMF (color, material, process) design - covering diverse materials, it aims to meet the core users' needs for texture and personalization. In the future, the company will also launch SKUs that meet the aesthetic preferences of female users and gradually expand to a wider general public.

In terms of business progress, Odyss plans to start KOL testing in North America in April this year, launch on the crowdfunding platform in June, and is expected to officially ship through its independent website from August to September. In addition, Odyss will accelerate its global layout, aiming to expand its business to the Chinese mainland, Japan, and European markets within two years and establish a complete channel and data link in China.

Facing market competition, Pan Yuyang believes that future AI hardware will not be an all - in - one like mobile phones, but will have different functions.

The necklace is the best form for capturing diet data. Although there are currently no direct competitors, it is in the stage from 0 to 1. If there are 3 - 5 direct competitors, they can jointly reduce supply - chain costs and educate the market.

In terms of the team, Pan Yuyang has worked on the Xiaoyi algorithm and HarmonyOS products at Huawei, and has also worked on the Coze smart glasses at ByteDance. The core team members mainly come from leading domestic and international Internet, hardware, and AI model companies. Currently, the team is in an expansion phase and is actively recruiting talents for key positions such as front - end and back - end development, algorithms, hardware R & D, and global marketing.