Service brief
HPE AI Services – Generative AI
Implementation
Develop your own Generative AI project and run it to address your
organization needs
What it looks like when industry practice is applied
Once you have identified the use case of reference driving the application success and related data that can come from
your organization or public sources, the next natural step is to experiment with the implementation, gain insights on
the value generated, and evolve and bring it to the core of your organization’s business.
Depending on the needs and available resources, you have a wide set of options to strategically select with HPE AI
Services – Generative AI Implementation. Should model generic abilities be enough for your case, HPE AI experts
implement the needed deployment model and optimize for inference, resulting in a ready-to-use solution to consume
and integrate with the existing processes or deploy as a stand-alone new application (see Figure 1).
When domain verticalization is needed, or complex contexts are provided as input to the system, effort is required
with a minor impact on resources and a high focus on quality data and business value. It is accomplished by applying
prompt engineering on top of inference tasks.
Moving forward with operationalization, a continuous feedback loop involving human evaluation is extremely important
for keeping the model performance at the initial level. For this reason, we apply Advanced Optimization Techniques
(AOT) developed by HPE AI expertise gained in years of experience working on production-ready AI solutions.
Why Generative AI? Why now?
Generative AI and the adoption of Large Language Models (LLM) represent one of the big artificial intelligence (AI) advancement waves.
The ability to let machine systems generate controlled outputs that are natively human readable tremendously simplifies their use,
accelerating market acceptance and experimentation enthusiasm. The key factors building the current success are:
• Novel model architectures such as transformers, leveraging the existing deep learning principles and available technology resources with
innovative structures to address existing areas such as natural language processing
• Open-source project proliferation, defining new job roles working on optimized frameworks to improve and optimize capability and on
model specialization with advanced optimization techniques and scalable fine-tuning approaches
• Large consumer availability, making cutting-edge discoveries available to anyone, including profiles with high creativity and less technical
background, accelerating the exploration of use cases
Finally, if there are specialized features that need to extend the model capabilities, it is required to build fine-tuned
versions from the model of choice involving a moderate use of computational resources. It is accomplished by applying
additional training either by updating the entire model or a portion of it, depending on how disruptive the desired
additional specialized feature is for the model.
We are open to organizations’ creativity, and should a completely new model architecture be required or the need to
use specific training sets arises with Generative AI and LLM design and build, we revisit the model architecture and
retrain on targeted and accordingly prepared data.
Figure 1. Generative AI modalities
What to expect
Our AI experts bring data science, machine learning (ML)
engineering, and ML Ops expertise to assess the business
expectations, align to technology requirements and
performance metrics, and prepare the required data in a
secure and proper format to be consumed by the models
(see Figure 2).
Data is the foundation of any AI project’s success. Dedicated
pipelines help ensure the smooth transition between source,
platforms, and output destination while loaders are used
to handle human-readable representations, which are
translated into machine‑readable formats keeping semantics
and related context.
Technology provides the needed resources to prepare and
run the deployed solution, optimized infrastructures, and
cloud‑native software platforms that help team collaboration
and agile development lifecycle.
Figure 2. Key engagement elements
What data, analytics, and AI advisory
Ecosystem
and professional can provide
Open-source projects and partner solutions are introduced when needed to
• Services — Expertise to explore,
experiment, and evolve your AI solutions
accomplish the business objective. HPE AI experts integrate and enhance the
ecosystem offerings, building and deploying a unique solution functional to
• HPE GreenLake edge‑to‑cloud platform
delivering AI, ML and data analytics, your desired purposes. Some examples of options are:
high‑performance computing (HPC) • Pachyderm, HPE Machine Learning Data Management Software
as‑a‑service on‑premises, in the cloud, or
in a colocation facility near you • Determined AI, HPE Machine Learning Development Environment
• HPE Software — Uncover hidden
• Aleph Alpha, Luminous is the collection of LLM from the EU-based startup
insights with data pipelining and
versioning, and help data scientists • Hugging Face, Model Hub shares democratized access to models and data
collaborate, build more accurate ML
models, and train them faster
• HPE compute and storage — Modern,
HPC solutions for GPU, data‑intensive
workloads, edge/IoT analytics, and
secure data management solutions
• Hewlett Packard Labs — Innovations
such as memory‑driven computing and
cookbooks for DL workloads based on
extensive benchmarking
• Partner ecosystem — Technology
and cloud partners offering data and
analytics solutions
Learn more at
HPE AI Services – Generative AI Implementation
HPE.com/greenlake Chat now (sales)
© Copyright 2023 Hewlett Packard Enterprise Development LP. The information contained herein is subject to change
without notice. The only warranties for Hewlett Packard Enterprise products and services are set forth in the express warranty
statements accompanying such products and services. Nothing herein should be construed as constituting an additional warranty.
Hewlett Packard Enterprise shall not be liable for technical or editorial errors or omissions contained herein.
a00134579ENW
