Bridging Academia and the Global Market: Groundbreaking AI and Deep-Tech Innovations Unveiled
At the prestigious SusHi Tech Tokyo arena, the 2nd National Network Conference (NINEJP in SusHi Tech Tokyo 2026) successfully gathered nine researchers and university-spun innovation teams. Supported by a nation-wide platform designed to provide global branding, database integration for venture capital, and commercialization support, these visionaries pitched deep-tech solutions ranging from real-time agricultural spectroscopy to medical imaging for rare cancers. Evaluated by an international panel of venture capitalists and incubation experts, the event highlighted the critical phase of translating cutting-edge laboratory research into viable, scalable market products. The grand prize was awarded to Science Tokyo's "Small Data AI" technology, which bypasses the massive data bottleneck to democratize cancer diagnostics.
Event Overview & Evaluation Framework
Master of Ceremonies (MC): Masaru Nagura (FoundersNation)
Pitch Format: 6-minute presentation followed by 3 minutes of interactive Q&A
The Prestigious Judging Panel
Ben Mumby-Croft – Director of Entrepreneurship, Imperial College London
Junji Takaoka – Incubation Partner, UTokyoIPC
Miwa Seki – General Partner, M Power Partners
Naoki Kamimaeda – Senior Advisor, Matsuo Institute & Venture Partner, IQ Capital
Michiko Kohyama – Founder, Startup DEI Promotion Council (Chief Presenter)
Evaluation Criteria
Technology Novelty & Advantages: Scientific and technical novelty in data utilization, roadmap to real-world deployment, and data acquisition/IP strategy.
Market & Social Impact: Potential market size, competitive advantages, and transformative value for society.
Commercialization Feasibility: Feasibility of the business model, funding plan, and strategic partnerships.
The Nine Pitch Profiles
1. Misato Itoh – HSFC / Obihiro University of Agriculture and Veterinary Medicine
Theme: Accelerating Livestock Field Diagnostics using Portable Spectroscopy and AI
The Problem:
Modern farmers face intense pressures from rising feed costs, labor shortages, and climate instability. Feed quality directly impacts livestock productivity and health. However, traditional feed testing requires sending samples to physical laboratories, taking 1 to 2 weeks and costing 6,000 JPY per sample, leading to delayed decision-making and potential losses up to 10 million JPY.
The Solution:
A handheld spectrometer paired with an AI analysis platform. By measuring grass or soil directly in the field, the AI instantly estimates nutrient parameters. This reduces the diagnostic time from weeks to minutes, and slashes the cost per sample to just 500 JPY.
Target Market & Roadmap:
Primary Target: Small-to-medium dairy farmers and agricultural cooperatives in Hokkaido (over 530,000 hectares of grassland).
0–12 Months: Expand POC to over 2,000 physical samples.
12–24 Months: Launch beta testing in partnership with JA (Japan Agricultural Cooperatives) and feed manufacturers.
24+ Months: Scale the service across the Hokkaido prefecture.
2. Takafumi Kubota – MASP (Michinoku Academia Startup Platform) / Tohoku University
Theme: POP (Personalized Medicine Platform): AI-Powered Genetic Testing & Drug Matching for Epilepsy
The Problem:
Epilepsy affects 1 in 30 people worldwide. Crucially, 40% of patients are drug-resistant. Finding the correct medication from approximately 30 available options is currently a slow process of trial and error (a "treatment odyssey"), where each drug trial takes over a year, meaning patients can spend decades suffering from uncontrolled seizures and driving/employment restrictions.
The Solution:
An AI-driven genetic testing platform. It identifies genetic causes in 60% of drug-resistant epilepsy cases by integrating whole-genome sequencing with clinical data. Furthermore, its generative AI agent slashes clinical literature search times from 1 hour to just 10 seconds, automatically generating gene summaries, identifying contraindicated drugs, and recommending optimal therapies.
Business Model & Roadmap:
Commercialization: B2B clinical tests sold to hospitals for $2,000 per test, alongside offering high-value genomic/trial datasets to pharmaceutical companies starting at $10,000 up to $1 million.
Global Target: Global market valued at $17 billion (22 million patients). Initial Japan market is worth $150 million.
Funding & Expansion: Raising $500,000 for 10% equity at seed stage (targeting a $50 million valuation in 5 years). Already conducting collaborative research with Stanford University and planning clinical trials (recruiting 100 to 300 patients) to target FDA approval in the US.
3. Shin Higeta – IJIE / Utsunomiya University
Theme: AI Dynamic Monitoring of Urban Environments
The Problem:
Traditional traffic surveys in commercial and street areas are highly manual, relying on workers sitting on chairs with clicker counters. On national highways, heavy, specialized analog equipment is required. These fragmented, localized methods make cross-city comparison and long-term trend analysis highly inaccurate and vulnerable to weather fluctuations.
The Solution:
An AI system that screens standard security and surveillance camera footage to seamlessly monitor pedestrian and vehicle flow 24 hours a day, 360 days a year. The software processes visual data in real-time to track object types (people, cars), speed, direction, and volume, storing anonymized data in a secure cloud database.
Partnerships & IP:
Collaborating with major infrastructure and security corporations including Tokyo Electric Power Grid (TEPCO) and ALSOK.
Strict adherence to privacy laws by processing non-personal, aggregated visual counts rather than facial or private personal data. The intellectual property is secured via proprietary software algorithms and public database protections of national universities.
4. Kenji Suzuki – GTIE / Institute of Science Tokyo (Science Tokyo)
Theme: Ultralight Small Data AI for Medical Diagnostics (MTANN)
【Grand Prize Winner】
The Problem:
While medical AI has proven capable of saving 20% more breast cancer patients by improving mammography accuracy, it suffers from a massive data bottleneck. Conventional deep learning models require at least 100,000 annotated images to develop, which takes years and requires expensive GPU server setups. As a result, 200 rare cancers (affecting 400,000 patients annually) are left entirely without AI support due to data scarcity.
The Solution:
A patented, ultra-lightweight technology called Massive-Training Artificial Neural Network (MTANN). Instead of full-image training, MTANN processes images pixel-by-pixel using micro-patches. This allows the AI to train highly accurate diagnostic models using fewer than 100 cases. Because the model architecture has three or more orders of magnitude fewer parameters (being 2,000 to 20,000 times smaller than state-of-the-art models), it is super-efficient and runs locally on a standard tablet without needing a GPU server.
Clinical Validation & SaaS Model:
Successfully developed a diagnostic AI for an extremely rare cancer, synovial sarcoma, using only 22 clinical cases, achieving a landmark 84% diagnostic accuracy.
Go-to-Market: A low-cost subscription model (similar to Netflix) where individual clinics and doctors download the app on their tablets and pay per use.
Expansion: Establishing startups in both Japan and the US, initiating FDA approvals in collaboration with leading global clinical sites including Harvard Medical School.
5. Keisuke Fujii – Tongali / Nagoya University
Theme: Multi-Agent Sports AI: Converting Soccer Video into Tactical Decisions
The Problem:
While recording match footage is standard practice, turning video into actionable tactical data remains difficult. Amateur and semi-professional teams lack the budget and personnel for quantitative analysis, wasting hours on manual editing. Conversely, professional clubs have fragmented, siloed data sets from multiple commercial vendors.
The Solution:
A comprehensive, no-code AI platform that unifies video analysis into a single interface:
Vision Board: Automatically extracts player and ball coordinates from raw video.
Open Style Lab: Integrates heterogeneous sports data into a standardized, common format.
Agent Suite: Evaluates full-scene positioning of all players simultaneously and models alternative "what-if" tactical choices.
Business Model & Validation:
Built on the rigorous foundation of over 120 peer-reviewed research papers and developed in collaboration with the Japan Football Association (JFA).
Open Core Model: Offers a free, open-standard layer to foster rapid integration across the sports ecosystem, while monetizing a proprietary layer that provides automated workflow reporting and team-specific tactical templates.
Roadmap: Testing beta software with 30-40 domestic Japanese football clubs (amateur/semi-pro) before scaling internationally and expanding to other fluid team sports like basketball and handball.
6. Chiaki Matsuzaki – TeSH / Ishikawa Prefectural University
Theme: Supercomputer-Driven Quality Design of Highly Stable Enzymes for Bio-Manufacturing
The Problem:
Enzymes are invaluable biocatalysts used to manufacture pharmaceuticals, foods, and biofuels. Japan has some of the world's deep-rooted expertise in enzyme research. However, natural enzymes are fragile, unstable, and easily degraded under harsh industrial temperatures and chemical conditions. Consequently, only 1% of discovered enzymes can currently be applied in industrial production.
The Solution:
A supercomputing-driven simulation platform that utilizes molecular modeling to stabilize enzymes. By introducing customized interatomic linkages (represented by simulated chemical bonds), the platform calculates the optimal stabilizing energies. This structural design drastically reduces the time and cost of physical "trial-and-error" wet-lab experiments.
Commercial Validation & Business Model:
Real-world Impact: Already deployed 3 customized stable enzymes into commercial use, generating an immediate $3 million annual sales increase for a manufacturing client.
Accuracy Target: Currently operating at 50% simulation accuracy, the platform is integrating physical lab feedback loop data to achieve 100% predictive design accuracy within 2 years.
Model: Monopolizing on a multi-tiered revenue framework consisting of upfront development fees, milestone payments, and long-term licensing royalties.
7. Ken Umeno – KSAC / Kyoto University
Theme: Postero-Xaos: Integrating Cyber, Disaster Resilience, and Finance with 256-bit Post-Quantum Chaos Encryption
The Problem:
Traditional earthquake warning systems only trigger after seismic waves are physically detected, giving citizens seconds to react. Additionally, scientific anomalies (pre-seismic electromagnetic signals) have historically struggled with trust and validation because there has been no secure way to prove exactly when an atmospheric observation was recorded before an event occurs.
The Solution:
A paradigm-shifting pre-seismic risk escalation platform. This technology monitors space weather and ionospheric anomalies that occur when tectonic stress causes massive electromagnetic disturbances prior to a rupture. Critically, to establish operational trust, the technology implements an immutable architecture:
Observation & Time Synchronization: Cryptographically secured time stamps.
Trust Layer: Utilizes Post-Quantum Cryptography (PQC) and tamper-resistant digital timestamps to mathematically verify the exact time and location of anomalous signals.
Real-World Application & Validation:
Analyzed data from the 2024 Noto Peninsula Earthquake, revealing anomalous ionospheric disturbances approximately 9 hours prior to the main shock.
Partnerships: Collaborating with local governments (Gifu Prefecture) and mission-critical transit operators (Shinkansen bullet trains).
Roadmap: 0-3 Months: Build the data-governance consortium; 3-9 Months: Integrate receiver datasets; 9-15 Months: Deploy a validated, tamper-resistant timestamped prototype.
8. Takafumi Shimizu – PSI / Shimane University
Theme: Non-Invasive Thermal Performance Evaluation & Building Energy Optimization
The Problem:
Japan faces a severe housing crisis with over 7+ million vacant homes, many of which suffer from poor thermal insulation and high heat escape through windows and walls. However, evaluating a building's thermal performance conventionally requires specialized human labor, extensive diagnostic equipment, and weeks of continuous measurements—making inspections highly expensive and cost-prohibitive for average buyers or renovators.
The Solution:
A non-invasive, AI-powered building performance evaluation platform. Instead of heavy diagnostic rigs, the system deploys simple, low-cost temperature and humidity sensors in target rooms, adjacent spaces, and outdoors. By applying specialized AI models, the platform calculates room-by-room insulation and air-tightness dynamics with a stunning 95% accuracy.
Go-to-Market Strategy:
Primary Target: Targeting the massive 1-trillion-yen housing renovation market, specifically focusing on housing manufacturers, architectural design offices, and local municipalities managing vacant properties.
International Footprint: A domestic patent has been filed, and the research team is already collaborating on global datasets with research institutions across Italy, the Netherlands, Germany, the United States, and Australia.
9. Masafumi Inomata – PARKS / Oita University
Theme: Smart OpeRec: AI-Driven Surgical Video Analysis and Auto-Documentation
The Problem:
Surgeons worldwide are severely overworked. After performing highly complex, hours-long surgeries, they are legally required to manually write detailed surgical reports. This paperwork relying on post-operative memory is time-consuming, prone to errors, and detracts from active patient care, while declining hospital work environments heighten the risk of medical lawsuits.
The Solution:
Smart OpeRec, a standalone AI software that processes high-definition surgical videos (endoscopic or laparoscopic) in real-time. By applying visual AI, medical decision modeling, and advanced clinical semantic processing, the system automatically generates structured, standardized medical records and clinical reports.
Clinical Value & Business Strategy:
Workforce Impact: Reduces administrative workloads by 1,000 hours per year per hospital, dramatically reducing costs while optimizing surgical team schedules.
Medical Safety: Provides objective, real-time data that supports data-driven risk prediction for post-operative complications and provides robust legal protection for hospitals through precise, audited records.
Roadmap: Validated using laparoscopic cholecystectomy (gallbladder removal, the world's most common laparoscopic surgery). Initiating clinical trials and joint research in 2026, aiming to expand through OEM partnerships first across Asia and then globally.
The Grand Finale & Judicial Insights
Following intense deliberation among the five expert judges, Prof. Kenji Suzuki (Institute of Science Tokyo / GTIE Platform) was awarded the Grand Prize for his revolutionary "Small Data AI" medical diagnostic platform.
In her concluding remarks on behalf of the panel, Michiko Kohyama emphasized the immense courage and talent demonstrated by all nine researchers in bridging academic science with the commercial market. She noted:
"The hardest space is translation—turning knowledge into real-world impact. In the market, it is no longer about whether your technology works in a controlled environment, but whether it can survive out in the real world. Do not underestimate the challenges ahead; engage early with your actual users. Success requires a diverse team that excels in business development, execution, and productization."
The Grand Prize winner received a trophy, active international conference and acceleration program support from the NINEJP platform, and a 6-month paid subscription to "Startup DB" provided by Four Startups.
Conclusion: The Future of Japanese Deep-Tech Commercialization
the 2nd National Network Conference (NINEJP in SusHi Tech Tokyo 2026) at SusHi Tech Tokyo showcased the immense potential of Japanese academic research when paired with robust market translation strategies. Historically, Japanese universities have been repositories of world-class science, yet they have struggled with commercializing these breakthroughs internationally due to language barriers, lack of early-stage venture funding, and a deficit in commercial business-development personnel.
Events like NINEJP and platforms like SusHi Tech Tokyo are turning the tide by:
Dismantling the Data Bottleneck: Solutions such as Science Tokyo's "Small Data AI" (MTANN) prove that Japanese startups can compete globally in AI by optimizing algorithmic efficiency rather than brute-forcing massive datasets.
Promoting Real-World Deployment: From agriculture in Hokkaido to housing insulation in Shimane, the pitches demonstrated a strong, practical commitment to immediate regional and societal challenges.
Fostering Cross-Disciplinary Teams: As highlighted by the judges, the successful translation of "lab to market" requires bridging the gap between scientific genius and commercial execution.
Ultimately, the event underscored a vital shift in the Japanese innovation ecosystem—moving away from closed corporate laboratories toward dynamic, university-spun deep-tech enterprises designed for the global stage from day one.
