TECH OFFERS

Despite the remarkable achievements of large language models (LLMs) in various tasks, there remains a linguistic bias that favors high-resource languages, such as English, often at the expense of low-resource and regional languages. To address this imbalance, we introduce SeaLLMs, an innovative series of language models that specifically focuses on Southeast Asian(SEA) languages. SeaLLMs are built upon the Llama-2 model and further advanced through continued pre-training with an extended vocabulary, specialized instruction and alignment tuning to better capture the intricacies of regional languages. This allows them to respect and reflect local cultural norms, customs, stylistic preferences, and legal considerations. Highlights: The models' attunement to local norms and legal stipulations—validated by human evaluations—establishes SeaLLMs as not only a technical breakthrough but also a socially responsiveinnovation. SeaLLM-13b models exhibit superior performance across a wide spectrum of linguistic tasks and assistant-style instruction-following capabilities relative to comparable open-source models. SeaLLMs outperform mainstream commercialized models for some tasks in non-Latin languages spoken in the region, meanwhile, SeaLLMs are efficient, faster, and cost-effective compared to commercialized models. The SeaLLMs went supervised finetuning (SFT) and specialized self-preferencing alignment usinga mix of public instruction data and a small number of queries used by SEA language native speakers in natural settings, which adapt to the local cultural norms, customs, styles and laws inthese areas. SeaLLM-13b models exhibit superior performance across a wide spectrum of linguistic tasks and assistant-style instruction-following capabilities relative to comparable open source models. Moreover, they also outperform other mainstream commercialized models in tasks involving very low-resource non-Latin languages spoken in the region, such as Thai, Khmer, Lao,and Burmese. Training Process Our pre-training data consists of more balanced mix of unlabeled free-text data across all SEA languages. We conduct pre-training in multiple stages. Each stage serves a different specific objective and involves dynamic control of (unsupervised and supervised) data mixture, as well as data specification and categorization. We also employ novel sequence construction and masking techniques during these stages.Our supervised finetuning (SFT) data consists of many categories. The largest and most dominantof them are public and open-source. As the aforementioned are English only, we employed several established automatic techniques to gather more instruction data for SEA languages through synthetic means. For a small number of SFT data, we engaged native speakers to vet, verify and modify SFT responses so that they adapt to the local cultural customs, norms, and laws. We also adopted safety tuning with data for each of these SEA countries, which helps to address many culturally and legally sensitive topics more appropriately - such tuning data tend to be ignored, or may even appear in conflict with the safety-tuning data of other mainstream models. Therefore, we believe that our models are more local-friendly and abide by local rules to a higher degree. We conduct SFT with a relatively balanced mix of SFT data from different categories. We make use of the system prompt during training, as we found it helps induce a prior which conditions the model to a behavioral distribution that focuses on safety and usefulness.   Through rigorous pre-training enhancements and culturally tailored fine-tuning processes,SeaLLMs have demonstrated exceptional proficiency in language understanding and generation tasks, challenging the performance of dominant commercial players in SEA languages, especially non-Latin ones. The models’ attunement to local norms and legal stipulations—validated by human evaluations—establishes SeaLLMs as not only a technical breakthrough but a socially responsive innovation, poised to democratize access to high-quality AI language tools across linguistically diverse regions. This work lays a foundation for further research into language models that respect and uphold the rich tapestry of human languages and cultures, ultimately driving the AI community towards a more inclusive future. One of the most reliable ways to compare chatbot models is peer comparison. With the help ofnative speakers, we built an instruction test set, called Sea-bench that focuses on various aspects expected in a user-facing chatbot, namely: (1) task-solving (e.g. translation & comprehension), (2)math-reasoning (e.g., math and logical reasoning questions), (3) general-instruction (e.g.,instructions in general domains), (4) natural-questions (e.g., questions about local context often written informally), and (5) safety- related questions. The test set also covers all languages that we are concerned with. AI model candidates' responses to the test set's instructions may be judged and compared by human evaluators or more powerful large and commercialized AI models to derive a reliable performance metric. Through this process, we demonstrate that our SeaLLM-13b model is able to perform on-par or supasses other open-source or private state-of-the-art models across many linguistic and writing tasks. Infocomm, Artificial Intelligence

As rapid global warming accelerates, the need for increased sustainability efforts has become a critical societal challenge. While individual lifestyle changes can contribute, their impact remains limited without broader systemic shifts. This places significant pressure on industries, particularly the fashion & textiles sector, a major contributor to climate change responsible for 10% of global greenhouse gas emissions. Decarbonising this industry is therefore crucial to achieving a sustainable future. This technology enables textiles and fabrics to remove carbon dioxide (CO2) from air. The patent-pending material functionalises textiles to capture CO2 present in air which is sequestered into a harmless mineral during the laundering process. The resultant mineral which is environmentally safe is then washed away, leaving the textile recharged to remove CO2 once more. With this technology, decarbonisation of the textiles industry can be achieved through the decentralised action of consumers utlising functionalised carbon removing products. The technology owner is interested in working with interested companies in the fashion industry value chain to test-bed this new material for carbon removing apparel and fabrics. The technology is formulated and provided in a liquid formulation, to be a drop-in process where it is embedded in textiles during the “finishing stage” (last step) of a textile mill. Some features of the carbon removing technology include: Continual recharging of functionalised textiles through normal laundering process Forms a stable and environmentally friendly mineral upon sequestration of CO2 by regular detergent Lasts at least 10 washing cycles Can be embedded with standard finishing equipment (particularly at the padding and stenting steps) Currently optimised for cellulose based textiles but proof of concept has demonstrated polyester, polyamide, wool and blends thereof This technology has been designed for textiles – both for apparel and functional fabrics. It can also be considered for non-woven materials as well as for other applications such as coatings. Facing immense pressure to reduce its environmental impact, the fashion and textiles industry, a major contributor to global warming, seeks sustainable solutions that don't disrupt its fast-paced production. With an addressable global market of US$227 billion for textiles, this innovative technology offers a solution to textile manufacturers to reduce the industry’s carbon footprint. This empowers consumers to become active participants in combating climate change, simply by choosing clothes made with this technology. Offers a proprietary, environmentally safe carbon removal solution for textile industry Continual usage of the functionalised textiles – textiles are rechargagle to remove CO2 multiple times Does not require the adoption of new machinery or processes for its implementation carbon dioxide removal, textile, additive, carbon removal, fabric, decarbonisation, fashion, clothing, materials, mineral, functionalisation, sustainable, sustainability, apparel Materials, Nano Materials, Chemicals, Inorganic, Additives, Sustainability, Low Carbon Economy

3D scanning by employing technologies like LiDAR, laser scans, TOF cameras and photogrammetry is an essential step in the process of making digital twins for the construction and built space sectors. This data is then meticulously processed, often manually, to form the 3D models for integration with Building Information Modeling (BIM) platforms and creation of accurate digital twins. The 3D models, by themselves or in conjunction with the realised Digital Twin, also help in different planning and monitoring tasks during the entire lifecycle of a building from construction to maintenance.  Obtaining the 3D cloud data from scanning and its conversion to a model is an involved process. The technology presented here eases both these processes by providing diverse options for scanning and enabling AI assisted conversion of the data obtained  to a 3D model capable of being used with BIM. The technology is compatible with multiple third party scanning solutions and also provides some native options - - mm level one shot scan using stationary laser scanner with a scan time of a few minutes. - cm level mobile LiDAR based scan. - cm level TOF flight based fast scanning. BIM conversion time is dependent on the kind of scan perfomed. For benchmarking, a test done using  THETA and BLK2Go where BLK2Go was used to walk around and scan the site while THETA was used to take necessary pictures, required conversion time of ~10 days including manual intervention, for a unit 2500 sqm in size at LOD 100. A further scanning option, using a handheld scanner requiring scans 5m apart and with a scan time of 1 second is also planned. This aims to remove the need of specialized personnel to conduct the scans. The solution also provides a model hosting option with GUI which gets automatically updated with new scan data for the same location regardless of the technology used. The previous data is still preserved and is available for review. Possible applications for the solution include - - Faster and more accurate site surveys with lesser manpower requirement. - Identification of available space at planned construction site and simulation of delivery routes, assesment of safety and workability. - Progress monitoring by using data in comparison to BIM along with assessing material requirements. - Intuitive management analysis and reduction of building lifecycle costs by monitoring parameters like human flow, energy requirements, equipment use amongst others. (Digital Twin) - Simulation of processes and optimization of production line efficiency. The technology uses AI to replace manual alignment of point cloud data and generate spacial linkages automatically thus saving substantial amount of time over current processes. The technology also provides option for an end to end solution encompassing scanning, data processing, digital twin generation and utilization. The generated results can also be hosted on a web platform which allows extension of use cases such as additional AI based solutions by third party collaborators. Infocomm, Artificial Intelligence, Green Building, Sensor, Network, Building Control & Optimisation, Smart Cities

The technology presented is an Artificial Intelligence (AI) model developed to extract essential information from scanned medical certificates. The trained model can extract pertinent details from medical certificates issued locally in Singapore and can help companies streamline their medical leave management  process by automating the approval of medical leave requests. The extracted details can also help in seamless integration with a company's existing workflow. The technology enables prompt and precise handling of leave requests and thus reduces administrative workload, processing time and errors introduced due to manual entry. The trained AI model recognizes terms and entities from scanned medical certificates. This includes but is not limited to - Clinic name Clinic address Clinic telephone number Patient name Start date of medical leave End date of medical leave Duration of medical leave The Name Entity Recognizer (NER) model is trained based on an open-source library and can be integrated with the existing workflow or system to automate the extraction of information for approval or recording purposes. The model, in its current state, is trained on a diverse dataset of medical certificates issued in Singapore and is suitable for application in systems providing Document Management and Human Resource solutions. The application of the model will particularly be useful for - Companies looking to automate their medical leave processing or application workflow. Insurance providers. Vendors specialising in Document Management, HR software solution, Payroll, and Attendance solutions.  The model can be integrated into their existing solution to value add in the processing of medicate certificates. The model is implemented using Natural Language Processing and deals with the domain of Named Entity Recognition. It has been trained using a diverse dataset of medical certificates issued in Singapore and is able to recognize entries of interest automatically from a scanned copy of the document. The model is able to take in the variation of formats, prints and naming of the entries and provide a recognizable input to the software systems making use of it. Medical Leave, Documentation Management, Named Entity Model, Human Resource Software, Medical Leave Automation Infocomm, Artificial Intelligence

This biotechnology pertains to a modular cloud-based bioprocessing system designed to streamline and enhance the cultivation and analysis of biological cultures. Addressing the complexities and constraints of traditional bioprocessing, this technology simplifies operations, making advanced bioprocessing tools accessible to a broader range of users. It has shown its versatility across various segments including educational institutions, research labs, biotech and bio-manufacturing companies and even within the food service industry, providing an efficient, flexible, affordable and scalable solution for growing biological cultures. The system comprises a base bioreactor unit with multiple add-on modules, including a multitude of environmental sensors, linear peristaltic pumps, compressed gas flow regulators, and novel stirrers & boosters. It is also controlled by a propietary cloud-based software, which provides a number of benefits for device management. This allows the user to access the bioreactor from anywhere, monitor experiments in real-time, and receive alerts for any errors. The modularity provides flexibility in the co-development of various bio-manufacturing applications, especially in streamlining production.      This technology is applicable in industries ranging from bio-medical to bio-pharma to food tech to environment tech. It serves as a foundation for products like vaccines, metabolites, cultured meats, fermented foods, biofuels, adjuvants, microbial inoculants, etc. It is especially relevant for R&D departments and educational programs focused on biotechnology and life sciences. The technology surpasses current market offerings with its modular design, ease of use, flexibility and affordability, enabling users to customize their setup according to their needs without extensive training or investment, thus democratizing advanced bioprocessing. Bioprocessing, Modular Bioreactor, Cell Culture, Microbial Cultivation, Synthetic Biology Culture, Scalable Biotechnology, Biotech Education, Advanced Biomanufacturing, Cultivated Meats, Alternative Proteins, Bio-Medical, Bio-Pharma, Fermentation, Precision Fermentation, Traditional Fermentation Foods, Processes

Near-infrared (NIR) light, part of the electromagnetic spectrum just beyond visible light, has various applications, particularly in vital sensing and food analysis. However, existing technologies for generating NIR light present certain limitations. Traditional halogen lamps can emit a continuous spectrum from visible to NIR wavelengths but pose challenges such as considerable heat generation, short lifetime, and difficulties in light distribution control. As a modern alternative, near-infrared LED arrays offer advantages such as no heat radiation and long lifespan. However, they are not suitable for applications requiring a wide wavelength range due to a lack of continuous output across the entire NIR spectrum. The wavelength intensity variation of NIR-LED arrays also affected the consistency of sensing and analysis. To overcome these challenges, the technology owner has developed a unique NIR phosphor as a heat-free light source with a wide spectrum range, enabling degradation-free analysis. Especially in food analysis, prolonged exposure to a halogen lamp may damage food. In addition, the long lifetime of this NIR source reduced the need for frequent replacements, leading to cost savings. Moreover, it can irradiate broadband NIR light from a single source, enabling easy light distribution control and wavelength axis alignment and reducing wavelength intensity variation within the irradiation plane. These advantages ensure consistency and accuracy in sensing and analytical applications. The technology owner is seeking R&D collaborations with industrial partners interested in integrating this advanced NIR light source into their applications. Compared to conventional near-infrared LED light sources, this NIR light source has a broader spectrum width, making it more suitable for spectroscopic measurements, especially those using multiple wavelengths. Key features of this technology are: Wideband spectrum: 450nm to 1000nm (over 1000nm is under development) Output: approximately 3W per module Adjustable spectral shape and light distribution angle to meet different needs Easy light focusing control and uniform wavelength intensity on the irradiation surfaces Long product life: 40,000 hours (LED chip guaranteed time), 40 times longer than typical lifespan of halogen lamps Minimal accuracy loss due to temperature changes: output drops by 8% only when the temperature increases from 25°C to 75°C, compared to a 27% drop of other products This unique near-infrared light source can be widely applied to night vision, non-contact vital sensing, food analysis, medical diagnosis, agricultural analysis, and other fields. Potential applications include (but are not limited to): Night vision: surveillance camera, traffic monitoring system, etc. Non-contact vital sensing: health monitoring (heart rate, oxygen saturation), self-health care, etc. NIR spectroscopy: foreign matter inspection, fruit and vegetable analysis, internal quality check, fresh food quality control, etc. Fluorescence imaging: endoscope, fundus camera, etc. Wide wavelength range and adjustable spectral shape No heat generation: enable degradation-free analysis Long product life: reduce maintenance frequency and costs Enhance the consistency and accuracy of sensing and analysis near-infrared spectroscopy, vital sensing, night vision, Light source Electronics, Lasers, Optics & Photonics, Healthcare, Diagnostics, Foods, Quality & Safety

Paper packaging is a versatile material used for a wide range of products. Its widespread adoption is due to its renewable and relatively low-cost resource along with environmental benefits such as recyclability and biodegradability. While paper packaging offers several advantages, some drawbacks of the material include porosity and the lack of barrier properties against moisture, oil, and grease. To overcome these limitations, conventional coatings such as polyethylene (PE) or polyfluoroalkyl substances (PFAS) have been employed to impart the required barrier protection. However, during the paper recycling process, it is difficult to repulp the coated paper due to several factors and results in reduced recyclability of such packaging materials. The technology on offer is a water-based coating formulation that can be applied onto paper packaging surfaces to act as a barrier against grease, liquid water, and water vapour. The coating imparts barrier protection functionalities, improving the paper’s resistance to grease, liquid water, and water vapor significantly. Use of bio-sourced constituents in the coating also improves product sustainability. As the coating’s constituents are repulpable, recyclability of the paper packaging can be achieved. With increasing awareness of reducing packaging waste, the deployment of this technology will offer companies a recyclable paper packaging with notable barrier properties. The technology owner is seeking for R&D co-development, test bedding and IP out licensing opportunities of this technology with interested companies. The water-based barrier coating technology has the following features: Consists of bio-sourced constituents to improve product sustainability Enables repulping of coated paper, largely improving recyclability of such packaging materials Improved barrier to water vapour transmission (WVTR) - WVTR value as low as 100 g/m2.day (based on ASTM E96) Improved liquid water resistance - Cobb60 value as low as 10 g/m2 (based on TAPPI T441) Improved grease resistance - a KIT rating as high as 12 (based on TAPPI T559) Easily applied by standard coating equipment Potential applications include (but are not limited to): Paper-based food packaging Paper boards, bags, and shipping sacks Products requiring enhanced barrier paper packaging Improves paper-based product recyclability while improving barrier properties of the paper Utilisation of bio-sourced constituents in coating formulation increases product sustainability Offers an alternative to PE and PFAS coated paper that are difficult to repulp coating, barrier, packaging, paper, water-based, recycling, recyclable, pulp, sustainability, sustainable, circular economy Chemicals, Coatings & Paints, Foods, Packaging & Storage, Organic, Bio-based, Sustainability, Circular Economy

Fragrance and deodorising sprays for home care, fabric care and pet care applications often suffer from inconsistent and shortlived performance. This Nano Delivery Technology encapsulates fragrances, essential oils and other odourous compounds into nano sized biodegradable capsules that can anchor themselves efficiently to fibres and hairs, while regulating the release of the encapsulated compounds over prolonged time periods. The encapsulation process takes place at room temperatures, using low energy methods, that preserves the integrity of the actives. The technology is designed as a ready-to-use adjuvant allowing manufacturers to nano encapsulate the actives independently and easily using their existing process and production equipments.  Reduces particle size of actives such as fragrances, essential oils and other odorous compounds into the nano scale Particle size of actives can be adjusted between 20nm to 200nm, depending on final usage requirements Encapsulation material is naturally derived and biodegradable Imparts consistent actives release over time Prolong effective time by 3 folds This technology owner is keen to explore the application of this Nano Delivery Technology to home care, fabric care, and pet care. Increases effectiveness of fragrances, essential oils and other odorous compounds Reduces reapplication frequencies  Encapsulation process takes place at room temperatures, using low energy methods Compatible with mainstream manufacturing processes and equipments Available as a ready-to-use adjuvant Patent pending  nanotechnology, nano, nanoencapsulation, encapsulation, micro encapsulation, emulsion, nanomaterials, nano material, fragrance, home fragrance, home care, laundry care, pet care, fragrance spray, deodorising, pet spray, controlled release, flavors, air freshener, odor, fabric freshener, scent, upholstery, odour, sanitiser, fabric Materials, Nano Materials, Chemicals, Flavours & Fragrances, Green Building, Indoor Environment Quality, Additives, Sustainability, Sustainable Living

The Reconfigurable Workspace Soft Gripper (RWSG) is a bio-inspired, pneumatically actuated, shape morphing soft robotic gripper that is capable of rapid reconfigurability. It features passive retractable nails, bi-directional foldable petals, and a flexible palm to adapt to various grasping and manipulation tasks and requirements. The ability to rapidly reconfigure allows the RWSG to grasp a wide range of large, thin, hard, delicate, and deformable objects. These capabilities make the RWSG a uniquely advantageous tool for high mix low volume manipulation and packing scenarios such as food assembly, packaging of groceries, and packing of consumer electronics.  The RWSG features retractable nails to help in precision grasping of small, thin, and high aspect ratio objects. An optimized bi-directional finger flap design allows its fingers to morph into scoop-like shapes to easily manipulate granular and semi liquid items such as grains, jelly, stews, curries or scrambled eggs. A multi-material palm design helps regulate the RWSG’s aperture to adapt for large or wide objects. The RWSG utilizes low, safe pressures (-80kPa to 60kPa) to switch between and operate the various grasping modes.  High mix low volume manipulation tasks for consumer goods, logistics, and food industries can benefit from advanced robotics to meet evolving demands in productivity, safety, and sustainability. These sectors often require manipulation and grasping capabilities that cannot be achieved by conventional robotics using rigid grippers or end-effectors. The RWSG can provide reliable and safe robotic handling of a wider range of objects in these challenging scenarios using its adaptive capabilities. With the ability of handling a wider range of objects, RWSG automation setups can help reduce changeover times (less or no tool changes required), improve safety (humans are not required any longer for manipulation in hazardous environments), and even contribute towards sustainability (less overall resources required).   The RWSG has a unique structure that allows robust and safe grasping of a wide range of large, thin, hard, delicate, granular, and deformable objects. Its structure is composed of food safe, hypoallergenic silicones that can tolerate both high and low temperatures. These unique features far surpass the capabilities of traditional rigid grippers and end-effectors. The RWSG can be seamlessly integrated with all major cooperative manipulators currently available in the market.  Soft Robotics, End Effector, Robotics, 3D Printing Electronics, Actuators, Infocomm, Robotics & Automation