Over 10 years we help companies reach their financial and branding goals. Engitech is a values-driven technology agency dedicated.

Gallery

Contacts

411 University St, Seattle, USA

engitech@oceanthemes.net

+1 -800-456-478-23

Twitter Facebook-f Pinterest-p Instagram

Challenge

Inaccurate and Inefficient Manual Document Review Hindered by Scalability & Hidden Insights

Client:
Provenio

Provenio needed a solution to

  • Automate anomaly detection: Accelerate review speed and reduce reliance on manual labor.
  • Improve accuracy: Minimize human error and ensure consistent anomaly identification.
  • Scale efficiently: Handle massive document volumes without compromising review quality.
  • Uncover hidden insights: Extract valuable patterns and trends for informed decision-making.

Enkefalos Technologies' Solution

LLM Model Building and Deployment:

  1. Data Acquisition and Preprocessing: We collaborated with Provenio to curate a substantial dataset of relevant documents, ensuring quality and diversity for model training.
  2. Model Selection and Pre-training: We chose a pre-trained LLM model suitable for anomaly detection and fine-tuned it on Provenio’s specific document domain and criteria.
  3. Customization and Training: We trained the LLM on the prepared dataset, focusing on identifying anomalies based on deviations in formatting, language patterns, and specific content criteria.
  4. Testing and Refinement: We rigorously tested the model’s performance, adjusting and retraining to optimize accuracy and reliability.
  5. Deployment and Integration: We seamlessly integrated the LLM model into the document management system, ensuring smooth data flow and efficient processing.

Human-in-the-Loop Web Application:

  1. Design and Development: We built a user-friendly web application for human analysts to review documents flagged by the LLM.
  2. Context and Highlighting: The app provided comprehensive context for each document and visually highlighted potential anomalies identified by the LLM.
  3. Confirmation and Refinement: Analysts could confirm the LLM’s findings, provide feedback, or refine the analysis for continuous improvement.
  4. Seamless Integration: The web app seamlessly integrates with the LLM model and document management system, ensuring a smooth review workflow.

Improved Document Review Process:

By combining the fine-tuned LLM model with the HITL web application achieved:

  • Faster Reviews: The LLM handled initial screening, significantly reducing human review time and workload.
  • Enhanced Accuracy: Consistent anomaly detection by the LLM minimized human error and improved overall accuracy.
  • Scalability: The automated system efficiently handled large document volumes, ensuring timely processing and analysis.

Deeper Insights: The LLM uncovered hidden patterns and trends, providing valuable data for informed decision-making.

"

Enkefalos are a proven and trusted partner of provenio.ai. Over the last 12+ months we have collaborated on solving real-business problems, using artificial intelligence and machine learning, and they have been a reliable extension to our team.

Augmenting our team with Enkefalos’ skilled resources has enabled us to scale, develop and enhance core elements of our AI platform.

We would like to offer our continued thanks and appreciation to the whole team who have contributed to our success

Ben Clannachan

Head of Product & Delivery
provenio.ai

Related Projects

Agents facing slow, manual quote generation and policy management processes, leading to inefficiencies, errors, and frustrated clients

LLM/

Translate