Use Case Intelligence RSS Search use cases Search
AWS Software / SaaS United States

Arcee AI

Document Processing

So Arcee AI needed to break down its computational tasks into smaller, less complex routines. As the startup prepared to go to market, it needed a way to reserve compute as required while optimizing price performance.

Read the original story Explore all use cases
97.94% to 1.6-hour training using AWS Trainium
Additionally, Arcee AI has reduced its ML model training time from 17 to 1.6 hours using AWS Trainium, cutting the training cost by 97.94 percent.
For example, it takes a 7-billion-parameter model about 20–30 seconds to complete an inference, helping customers gain faster insights and responses.
Reduced from 17 hours

Reported impact

Business outcome

Using AWS, Arcee AI is empowering enterprises to harness the full potential of domain-adapted language models and drive transformative innovation. In the future, the startup plans to add more features to its SLM adaptation system to help customers rapidly adapt and scale their ML needs in line with their broader business goals. Key reported metrics include reduction in training cost using AWS Trainium 97.94%; 97.94% to 1.6-hour training using AWS Trainium; to 1.6-hour training using AWS Trainium Reduced from 17 hours.

  • Productivity Improvement
  • Scalability

Solution

Learn how software startup Arcee AI accelerated innovation using Amazon EC2 Capacity Blocks for ML, AWS Inferentia2, and AWS Trainium. Artificial intelligence (AI) is rapidly advancing, compelling many startups to develop sophisticated models to tackle a variety of use cases. The larger and more complex the model, the more the GPUs that are required to train it. Arcee AI develops domain-adapted small language models (SLMs) to help enterprises perform specialized tasks, such as analyzing legal documents. It needs to have reliable access to GPUs to manage its compute-intensive workloads. When Arcee AI went to market, it needed a large amount of accelerated compute resources to train the many SLMs that it is continually developing. But because of ongoing GPU shortages, the startup lacked the compute that is required to effectively train its models and bring them into production.

Data flow

Source data enters the Arcee AI workflow through its application services, stored or indexed in Amazon OpenSearch Service, processed with Amazon EC2 Capacity Blocks for ML, AWS Inferentia2, AWS Trainium.

Solution architecture

9 components · 2 layers
  1. Compute
    • Amazon EC2 Capacity Blocks for ML Reserves accelerated compute capacity for ML training workloads.
    • AWS Inferentia2 Accelerates model inference with custom AWS silicon.
    • AWS Trainium Accelerates deep learning training with custom AWS silicon.
    • Amazon ECS Runs containerized workloads without self-managing orchestration control planes.
    • Amazon EC2 Trn1 Instances Runs high-throughput deep learning training jobs with AWS Trainium.
    • Amazon EC2 Inf2 Instances Runs cost-efficient large-scale inference with AWS Inferentia2.
    • Amazon SageMaker Supports model training, fine-tuning, evaluation, or managed inference for the customer workload.
    • Hugging Face Provides access to pretrained models that the customer adapts for domain use cases.
  2. Storage
    • Amazon OpenSearch Service Stores searchable indexes or vectors and serves low-latency search and analytics queries.

Architecture clues

  • In the future, the startup plans to add more features to its SLM adaptation system to help customers rapidly adapt and scale their ML needs in line with their broader business goals.
  • Jacob Solawetz Cofounder and Chief Technology Officer, Arcee AI Artificial intelligence (AI) is rapidly advancing, compelling many startups to develop sophisticated models to tackle a variety of use cases.
  • To help developers manage GPU reservations, the company built a web application that is hosted through an API architecture running on Amazon Elastic Container Service (Amazon ECS), a fully managed container orchestration
  • To quickly surface the information that they need to perform analytics, Arcee AI and its customers use a vector database that is powered by Amazon OpenSearch Service , which unlocks near real-time search, monitoring, and

Evidence from the source

97.94% to 1.6-hour training using AWS Trainium
Additionally, Arcee AI has reduced its ML model training time from 17 to 1.6 hours using AWS Trainium, cutting the training cost by 97.94 percent.
For example, it takes a 7-billion-parameter model about 20–30 seconds to complete an inference, helping customers gain faster insights and responses.
Reduced from 17 hours
So Arcee AI needed to break down its computational tasks into smaller, less complex routines. As the startup prepared to go to market, it needed a way to reserve compute as required while optimizing price performance.
Using high-performance compute instances on AWS, Arcee AI has been able to efficiently train models with extensive datasets, including one PubMed dataset with 88 billion tokens.
Transparency Evidence & methodology 95/100 reference value · 90% confidence · 2 sources

Review how this reference was scored and trace each normalized claim back to its documented source.

01

Scoring methodology

How scores are calculated

Customer Reference Value measures usefulness as a customer reference. Evidence Confidence measures the quality and traceability of available evidence, not independent verification.

Customer reference value95/100
  • Problem clarity20/20
    Measures whether the customer's operational problem is specific enough to compare.
  • Implementation depth20/25
    Rewards named components, data flow, integrations, and architecture detail.
  • Outcome specificity20/20
    Measures whether the reported benefit is concrete rather than a generic claim.
  • Quantified impact25/25
    6 metrics describe measurable impact.
  • Deployment scale10/10
    Rewards evidence of a live or scaled deployment and meaningful integration scope.
Evidence confidence90%
  • Source authority20/25
    The main evidence is an official vendor-published customer story.
  • Claim traceability23/25
    11 of 12 key fields link to a named source.
  • Direct evidence20/20
    6 direct source quotes retained.
  • Date provenance8/10
    The publication date comes from publisher-controlled metadata.
  • Source transparency9/10
    2 source references document publisher, URL, authority, retrieval, and supported fields.
  • Implementation specificity10/10
    Rewards source-backed components, data flow, quantified impact, and a specific outcome.
02

Traceability

Sources & provenance

Field coverage shows which facts each source supports. Dates are shown only when documented by the source.

  1. AWS

    Arcee AI customer story

    First-party
    https://aws.amazon.com/solutions/case-studies/arcee-ai-case-study/
    Source type
    Vendor-published customer story
    Published
    Aug 19, 2024
    Supports
    • Customer
    • Business problem
    • Solution
    • Products
    • Solution components
    • Data flow
    • Integrations
    • technical area
    • use case category
    • Business outcome
    • outcome category
    • Metrics
    • Architecture
    • evidence quotes
    • Publication date
  2. aVenture

    Arcee AI - aVenture Company Research

    Independent secondary
    https://aventure.vc/companies/arcee-ai

    The company profile states 'Operating headquarters' in 'Miami, FL, US'.

    Source type
    Structured company profile
    Supports
    • Region

Date evidence

Not separately sourced Industry, Company size. Treat these normalized attributes as lower-confidence until a dedicated source is added.

Related

Similar use cases

See all like this
AWS Jan 6, 2025 Smokeball Software / SaaS Global While digitalization had made significant progress, the company identified that innovative law firms required more sophisticated tools to optimize operations and enhance client service. Due to challenges with… Founded in 2012, Smokeball serves over 6,000 law firms with 25,000 daily users across Australia, the UK, and… Document Processing AWS Lambda · Amazon Bedrock · Amazon EKS AWS Dec 8, 2024 meeting.ai Software / SaaS Indonesia # meeting.ai Optimizes Meeting Productivity with Amazon Bedrock and Claude 3 Explore how meeting.ai is revolutionizing meeting productivity Explore how meeting.ai , an innovative Indonesian startup, is leveraging the… Document ProcessingEnterprise Knowledge Search Amazon Bedrock · Amazon Elastic Kubernetes Service · Amazon Managed Streaming for Apache Kafka AWS Nov 24, 2024 GTxcel Software / SaaS United States In the digital age, it comes as no surprise that printed documents are steadily declining. Once the primary method of consuming content, many of us now favor the convenience of accessing a rich array of media from a… 10 years Document Processing AWS Lambda · Amazon Bedrock · Amazon S3