CAP Vista Accelerator: Solicitation 3.0 (Multiple Technologies)

Application Portal: Cap Vista Accelerator

Solicitation Page: Cap Vista Accelerator - Solicitation 3.0

Application Deadline: April 1st, 2025

Inviting global companies and startups to address challenges in areas of artificial intelligence, communications, cyber security, cognitive performance etc.

Cap Vista Pte Ltd in Singapore is hosting a solicitation for international startups to complete and pitch their solutions for the following categories:

1. We are seeking novel approaches to enhance the selection process for pilots.

Artificial Intelligence could possibly assist companies in adopting novel approaches to enhance their candidate selection process to identify and engage potential candidates.

What We Are Looking For

We are seeking novel approaches to assess and identify potential candidates for the pilot vocation. Your approach should:

• Clearly explain the concept, critical attributes (for example: limited to perceptual ability, psychomotor, multitask performance) to identify, engage, and select the most-suitable pilots candidates.

Key Metrics

• Ease of implementation of your approach.

Your Technology Solution Should Demonstrate

• Novelty and effectiveness in your approach to select pilot candidates.
• Ability to attract and engage candidates, aged 18 to 25, to participate in a time-efficient and convenient selection process.
• Ability to collect psychometric attributes of multiple candidates effectively.

Trial Specifications

• Trial data collected should be secure and exportable to other formats such as Excel, SPSS, SAS, or R.
• The company shall work with end-users to analyse trial results and include trial report with useful insights.

What We Are Not Looking For

• Hardware solutions or dedicated apps.
• Questionnaires that lack elements of engagement and fun.
• Restrictive solutions that may not?reach out to a wide candidate pool.
• Solutions requiring long developmental time.

Option Cost

• Cost of longer-term engagement beyond the trial.
• Adaptation for use in other vocations’ selection process.

2. How might we enhance security for processing private data using public Graphics Processing Unit (GPU) infrastructure?

Context

Public GPU infrastructure could provide organisations with agility and flexibility in leveraging Artificial Intelligence (AI) and Machine Learning (ML) capabilities. However, there is a need to mitigate potential exposure to cyber-attacks on shared computing environments, and have effective security controls over data access and processing.

What We Are Looking For

• Novel ways to securely process private data on public GPU and computing resources.
• Solution may leverage hardware-based controls to mitigate software vulnerabilities.

Key Metrics

• Volume of data that could be processed in a given time.
• Robustness of cyber security features.
• User-friendliness, maintainability, and scalability.

Your Technology Solution Should Demonstrate

• Ability to ensure data security, while ensuring data owners maintain control without unauthorised access by GPU providers, data centre operators, or third parties and hackers.
• Solutions that can constantly ensure components are trustworthy and functioning correctly.
• Scalability to different computing environments without affecting the performance of AI/ML tasks.
• Ability to meet security standards.

Trial Specifications

• A scope to showcase tech capabilities over a specified trial duration.

What We Are Not Looking For

• Software-only security solution.

Option Cost

• Longer-term engagement beyond the proof of concept.

3. How might we ensure resilient and secure device-to-device communication during a mobile communication network disruption?

Context

When a mobile communication network capacity is disrupted, the consequences can be severe for emergency responders.

What We Are Looking For

We are looking for Android-compatible or Android-integrated software solutions to provide resilient and secure mesh node connectivity through mobile devices for device-to-device communication during a mobile communication network disruption.

Key Metrics

• Achievable throughput and latency.
• Ability to handle increase(s) in the number of mesh nodes and traffic load.
• Consistency and stability of mesh node setup and connectivity.
• Uniqueness, effectiveness, and key differentiation?of the solution.

Your Technology Solution Should Demonstrate

• Continuous communication, including transmission of data, video, and voice over the direct device-to-device mesh node connectivity.
• A software-based solution with the ability to integrate into Commercial-off-the-shelf (COTS) mobile Android-based devices.
• The ease of?setup and operation, as well as addition of new mesh nodes (i.e. mobile devices).

Trial Specifications

• The company is to propose a scope to showcase the tech capabilities and to collect the key metrics, as mentioned above.
• The trial should allow multiple users to gather the test metrics across indoor and outdoor environments as determined by us.
• We anticipate a trial of up to 6 months with support and on-site testing to last not more than a week.

What We Are Not Looking For

• Additional hardware as part of the setup is not preferred.

Option Cost

Cost based on the number?of units:

• < 1,000,
• 1,001 to 2,999, and
• > 2,999 units.

4. We are seeking innovative digital solutions to better track, manage, and recommend asset movement.

Context

Computer applications that can track asset movements and their localities and support the deconfliction of asset movement requests and locality availability. The application should be able to provide decision support recommendations to prompt the operator to move assets in and out of the localities based on predefined rules and events.

What We Are Looking For

Digital solutions that can optimise operator workflow, reduce conflicts, and enhance operational safety by:

• Introducing automation and decision support technologies to relieve cognitive workloads.
• Aggregating operator inputs to generate and update unified situational picture of the assets.

Key Metrics

• Effectiveness of the solution, for example, number of users or volume of manageable traffic.
• User-friendliness of the solution, including its usability and intuitiveness.

Your Technology Solution Should Demonstrate

• Ability to pre-empt operators of any conflicts, such as overlap in asset movement paths.
• Potential to manage more assets in larger areas.
• Potential to integrate with other operational systems to support data flow.

Trial Specifications

• On-premises solution to be used by operators for comparison with existing workflow.
• Trial duration to include operator training and follow-on tech support, where required.

What We Are Not Looking For

• Resource-intensive solutions that introduce more cognitive load to operators.
• Internet-of-Things asset tracking solutions.

Option Cost

To quote the following:

• Longer-term subscription of the solution, including maintenance and system upgrades.
• Integration with other systems.

5. We are seeking solutions to detect, recognise, and track objects easily with different sensor performance and viewpoints.

Context

Changes in sensor performance and viewpoints may require repeated algorithm refinements or developments to detect, recognise, and track objects.

What We Are Looking For

We are looking for a software that accurately recognises objects without additional development efforts in the different scenarios below:

• Sensor resolution of the object.
• Sensor’s look angle at the object.
• Sensor’s distance from the object.
• Using one or a combination of visual sensors.

Key Metrics

• Accuracy in object detection.
• Accuracy in object recognition.
• Computational time – should have real-time capability if possible.

Your Technology Solution Should Demonstrate

• Performance analysis, such as reliability, computation timing, and load.
• Limitations and trade-offs, if any.

Trial Specifications

• The company is to propose a scope to showcase the software capabilities and to collect key metrics, as mentioned above.
• The trial should include edge cases and true negative verification.
• The test set should contain sufficient variability.

What We Are Not Looking For

• Academic papers.
• Hardware solutions.

Option Cost

• Cost for deployment beyond trial.

6. We are seeking an on-premises software tool to extract accurate answers from a large repository of documents containing jargons, acronyms, hand-drawn images, graphs, charts, and tables.

Context

It is time-consuming to manually search and extract relevant information from a large repository of documents (e.g. ppt, pdf) with acronyms, hand-drawn images, graphs, charts, and tables.

What We Are Looking For

We are looking for an on-premise software tool that is (i) able to?generate and cite accurate and relevant AI-crafted?answers from a large user-uploaded repository of documents (e.g. ppt, pdf) with jargons, acronyms, hand-drawn images, graphs, charts, and tables based on users’ submitted text queries and? (ii) low-code and customizable to allow users to change system prompts, and configure various models and settings.

Key Metrics

• Search accuracy and relevance for visual-rich documents.
• Contextual understanding and nuance capturing.
• User interface usability and intuitiveness.
• Low-code and customisable to allow users to tailor the software to different use-cases.

Your Technology Solution Should Demonstrate

Ability to make sense of the text and images in the documents and to answer queries within the local network with no internet connection.

Trial Specifications

The tool will be loaded onto our furnished GPU-enabled laptop for an approximate 6-month trial.

What We Are Not Looking For

• An application that requires significant pre-training.
• An application that requires internet connection.

Option Cost

• Cost for?on-premises deployment.

7. How might we detect unsafe behaviours or conditions without any prior trained data?

Context

Artificial Intelligence (AI) has been used in video surveillance but its application could be limited by its need for context-specific training data.

What We Are Looking For

• A video analysis platform that can automatically detect unsafe behaviours or conditions in real-time without predefined examples.
• A context-aware AI, capable of interpreting safety policies and applying them dynamically.

Key Metrics

• Accuracy and reliability in detection and even classification of unsafe behaviours or conditions.
• Ability to give real-time alerts.

Your Technology Solution Should Demonstrate

• Ability to recognise actions in surveillance videos without the need for labelled training data.

Trial Specifications

• The company is to propose a scope to showcase the software capabilities using its own camera(s) and other necessary setup to collect key metrics as mentioned above, during an onsite trial.

What We Are Not Looking For

• An application that requires significant pre-training.
• An application that requires internet connection.
• Safety training workshops.

Option Cost

• Cost for?on-premises deployment.

8. How might we utilise cameras and computer vision to provide accurate geographical location for Unmanned Ground Vehicles (UGVs) and self-driving vehicles?

Context

Unmanned Ground Vehicles (UGVs) and self-driving vehicles need to accurately locate themselves in outdoor urban terrain, even when navigational systems such as Global Navigation Satellite System (GNSS) signals are intermittent.

What We Are Looking For

• Solutions?that provide accurate and reliable location coordinates with existing vehicle sensors such as onboard cameras.
• Solutions that enable vehicles to navigate seamlessly with intermittent/disrupted GNSS signals.

Key Metrics

• Accuracy in determining vehicle’s position.
• Latency or real-time processing.
• Ease of integration.

Your Technology Solution Should Demonstrate

• Utilisation of onboard cameras to pinpoint vehicle’s location on a preloaded map or provide the latitude and longitude of the vehicle.

Trial Specifications

• The company can use its own standard motorcar with mounted cameras for trials.
• The motorcar will be driven around in urban environment and tested for localisation accuracy (with intermittent/disrupted GNSS signals).

What We Are Not Looking For

• A system that requires extensive modifications to the vehicles.
• A system that requires additional active sensors (e.g. LIDAR, radar, etc.).

Option Cost

• Cost for deployment.