Startup Battlefield 200: nine Korean startups eye the US
Nine South Korean deep tech startups pitched in Seoul for a spot in Startup Battlefield 200 at Disrupt 2027. Voinosis, using AI on voice, won.
Nine early-stage South Korean startups presented their projects at DETO 2026, a technology conference in Seoul, to win a place in next year’s Startup Battlefield 200. They work in deep tech, meaning advanced scientific and engineering technologies, and all want to enter the US market as part of broader international expansion plans.1 Startup Battlefield 200 is TechCrunch’s annual competition in which hundreds of companies from around the world try to pitch to venture capital funds (investors in young companies) and the international tech community during the Disrupt conference. The Seoul challenge concerns the 2027 edition.1 The sectors covered by the nine companies are healthcare, financial technology, energy, semiconductors and artificial intelligence.1
Who won in Seoul and what Voinosis does
The competition was won by Voinosis, which develops an artificial intelligence to detect changes in brain health through voice. The tool analyses pitch, pauses, rhythm and response times, and each check lasts less than five minutes.1 The company says the technology has been tested in Greek, Japanese and English without needing separate training for each language. In a clinical study with 146 participants, it claims to have reached 88% accuracy in identifying possible cognitive problems.1 Voinosis targets hospitals, nursing homes, insurers and private consumers. It also plans to connect voice checks to cognitive training, assessments and care provided by healthcare professionals.1 The founder is Grace Shin, an ear surgeon for over 25 years. The startup is raising a Series A round and already has a South Korean pharmaceutical company among its strategic investors.1
The other eight startups
Awesome Lab, based in Seoul, produces ThermoFlux, a heating system that generates heat through water electrolysis and ionization. It can serve buildings, water supplies and industrial processes, and according to the company it reduces energy consumption and carbon emissions. In the US it plans to supply OEM modules (components for other manufacturers) and to develop products with boiler, water heater and air conditioning companies.1 Crosshub helps travellers and foreign visitors verify their identity and set up payment methods abroad. Its IDBlock platform uses self-sovereign identity (SSI) and zero-knowledge proof, techniques that allow someone to prove who they are without revealing unnecessary personal data.1 DeepFusion AI has developed an imaging system with 4D radar and other sensors for autonomous mobility, robotaxis, maritime surveillance and defence. It combines radiofrequency sensors, radar and lidar to detect objects, movement or distance even without a direct line of sight, in a single customisable unit.1 EFLOW builds motors that produce more torque in less space, for robots, drones and satellites, and actuators that control the movement of robotic joints. Otiton Medical makes two devices: a body thermometer and an instrument that captures images of the ear canal and analyses the eardrum. The company states that its AI was trained on about 55,000 eardrum images and that the device has received medical certification in South Korea.1 Riibotics says it can install robots in warehouses in a single day, simplifying infrastructure, mapping, system integration, task programming and safety controls. Testbank develops Solve, an app that turns printed school textbooks into interactive digital versions: the AI identifies the spaces for questions and answers in publishers’ PDFs, keeps the layout and corrects automatically.1 Ternell, a semiconductor startup from Ulsan founded in 2019 by KyungRok Kim, a former Samsung Electronics engineer, is developing a new type of memory to reduce the power consumption of AI chips. Its T-SRAM system adds a third state to the two of conventional memories. In tests on 28-nanometer chips, the company reports a capacity-per-area increase of between 33% and 102% and a drop in the energy needed to access data of between 55% and 71%, compared with traditional SRAM.1
What changes
The nine companies have very different products, from voice as a diagnostic tool to robot motors, and the declared common thread is the US market. The TechCrunch competition gives them a stage in front of international investors and observers.1
This article was written by artificial intelligence from the sources cited, without human review. Automatic newsroom by Redazia.
