Resources

MODELS

Leveraging best-in-class FMS & LLMs

Users have the option to use out-of-the-box pre-trained models from third parties, open source or proprietary models. This flexibility allows to start small, and progress to build larger solutions that precisely meet the complexities and budget of the task at hand.

Omnitive automatically transforms natural language users’ input into prompts. This completely avoids prompt engineering. 

Users can progress from zero-shot to few-shots and finally model fine-tuning. As needed, Omnitive will go on to gather one to few examples, process them and generate the most accurate response. At any time, users can query the performance of the model through natural language, iterating until the results are satisfactory.

Should zero and few-shot approaches fall short, Omnitive features base model fine-tuning for FMs and LLMs. This is done using natural language user input and a small model adapter. Omnitive leverages LoRA and QLoRA techniques, which are the latest research in model adaptation to reduce training time and cost. 

When serving inference models, Orkid is able to optimise the cost by merging the frozen FM, made of billions of parameters, with the small adapter created during fine-tuning. While data privacy is enforced, the cost of hosting a large model can be shared across, thereby significantly reducing individual bills. 

Three different sources:
① Customer curated
② Fully generated
③ Augmented

The best trade off in terms of quality and quantity is the Augmented training data. In this case the user starts with an out-of-the-box model to produce some inputs and outputs. Omnitive is then able to augment the rest of the dataset based on that small initial input.

In all cases, users do not need to produce a whole dataset by hand, even the initial seed dataset can be generated using natural language instructions through Omnitive Partner.

INTEGRATIONS

Seamless Integration, Shorter Time to Value

Information access solutions don’t live in isolation. They require integration with other down- and up-stream systems, such as RPAs, ERPs, Accounting, Document Management, or workflow tools. That is why Orkid offers several pre-built off-the-self integrations, enabling a shorter time to value and higher ROI.


Some of the most salient integrations available include IBM RPA, IBMx and Appian.

security

Safety, security and peace of mind

Safety security and peace of mind

Orkid has an ample experience dealing with private, as well as secret information for companies and government all around the globe. We’re fully committed to maintaining the highest standards when it comes to keeping customer data safe and private, while in compliance with the latest relevant policies.

To make sure this is the case, we’ve put together a solid set of policies, procedures, and controls. They help us make certain that your data remains confidential, unaltered, and available whenever you require it. Please refer to the following sections for additional detail.

Orkid’s Data Protection Officer (DPO) oversees the implementation and management of our security and privacy policies. The goal is to ensure that Orkid processes customer data, as well as the personal data of its staff, providers or any other individuals or business affiliated to Orkid in compliance with GDPR.

Orkid’s DPO is entrusted with performing data protection impact assessments, conducting audits, training staff, and monitoring ongoing compliance with GDPR, while acting as the focal point for the data protection supervisory authority.

For any additional information related to Orkid’s management of customer data and personal information, our security and privacy practices, or application of GDPR do not hesitate and reach our Data Protection Officer.

We’re on top of things when it comes to the technical and organizational side. Our internal controls and security procedures are state-of-the-art, following the best practices in the industry. Our aim is to ensure your data stays safe and sound – no accidental losses, no unauthorized access, and no tampering or shady business. These measures involve making sure our team members handling your data are dependable, keeping a tight grip on access, employing robust authentication methods, providing thorough training to our staff, frequently backing up data, having contingency plans in place, dealing with any incidents that might pop up, and placing constraints on data storage, printing, and disposal. Plus, we’re extra vigilant about safeguarding the devices that house your data.

These third-party certifications are recognized and accepted internationally representing proof of both, Information Security (ISO 27001) and Quality of Processes (ISO 9001). 

ISO 27001 certification provides evidence on how Orkid addresses information security risk management. It demonstrates to customers that we have in place all the relevant security systems, policies, and controls to ensure the integrity of sensitive information. 

ISO 9001 certification shows the importance an organization places on improving its customer-facing systems by focusing on quality. It demonstrates that Orkid has an effective Quality Management System (QMS) in place enabling us to dependably provide superior value-driven products and services. 

ISO 27001 and ISO 9001, while focusing on two different and important aspects of business, share the common goal of customer satisfaction, and ultimately improving organizational effectiveness and efficiency.

Orkid continues to build its cloud independent strategy.  From it’s historic roots partnering with Microsoft Azure across Asia, the majority of Orkid’s European customers in both public and private deployments are leveraging AWS powered solutions today. These global partners provide secure, extensive, and reliable global cloud infrastructure for hosting our customer data and applications. They all meet the highest security standards worldwide, supporting hundreds of security standards and compliance certifications such as PCI-DSS, HIPAA/HITECH, FedRAMP, GDPR, FIPS 140-2, and NIST 800-171. This ensures that our customers can easily meet compliance requirements no matter where they operate.
cloud deployment

Scalability, privacy and performance

Public or Private Cloud deployments are available to meet the specific scalability, privacy, and performance of your applications. Omnitive relies on Docker containers and Kubernetes, inclusive of advanced logging and monitoring, to give IT teams the necessary control over the infrastructure. Omnitive integrates with virtually any application through API REST.

Public Cloud

  • Virtually unlimited computing capacity & storage
  • Faster scalability
  • No additional operational costs
  • Lower infrastructure costs
  • Multi-tenant or single-tenant options available
  • Shorter time to value means faster ROI.

Private Cloud

  • Customer owned and managed.
  • Increased access control and resource customization
  • Higher levels of security and privacy
  • Unified governance and data access
  • Improved performance

Best Efforts

Omnitive MIDP use the latest and most powerful Generative AI algorithms. Depending on document quality, document classification and data extraction accuracies will vary, normally upwards of 90%. Best efforts gives direct access to the unsupervised outcome of the algorithms.

Omnitive MIDP supervision tool is made available to further further lift accuracies as appropriate.

TRADE OFF
Higher Speed at Lower Cost

BENEFITS
Flexibility to self-review outcomes
Orkid’s supervision tool
Faster outcomes
Lower price point per outcome

SLA

Recommended for customers that demand the highest level of accuracy and for mission critical applications. We will agree to a specific SLA in terms of accuracy and speed of delivery of outcomes. Orkid will deploy its team of data engineers to review and improve outcomes and meet the contractual SLA.

All outcomes delivered meet the agreed levels of accuracy.

TRADE OFF
Higher Value at Higher Cost 

BENEFITS
Guaranteed outcome quality
Consume and forget
Orkid’s data engineers in the loop
Higher quality per outcome

faq

Frequently asked questions

We offer our customers the option to use out-of-the-box pre-trained models from third parties open-source and proprietary sources. Large Language Models (LLM) are or can be pre-trained on vast Corporate data and are commonly referred to as Foundation Models (FMs). The FMs contains billions of parameters and variables to go into their actual makeup. A single FM can perform many different tasks not just one.

No. Omnitive automatically transforms user natural language input into prompts. This completely avoids prompt engineering.

Users can progress from zero-shot to few-shots and finally model fine-tuning.

Yes. Should zero and few-shot approaches fall short Omnitive features base model fine-tuning for FMs and LLMs.

Yes it is.

Yes Orkid is easy to integrate through API REST with virtually any application.

Yes we us Docker for creating sharing and running individual containers and Kubernetes for operating containerized applications at scale.

Yes Orkid counts with ample experience dealing with private as well as secret information for companies and government all around the globe.

We currently offer Best Efforts and SLA options.

Best efforts give direct access to the unsupervised outcome of the algorithms.

It greatly depends on document quality. With reasonable quality documents classification and data extraction accuracies will normally be upwards of 95%

Orkid’s supervision tool is made available to further lift accuracies as appropriate.

Download our FAQ Document