Saga Wisdom Unlocks AI-Powered Learning with Video Intelligence on AWS
At a glance
Saga Wisdom partnered with New Math Data to transform over 850 hours of “dark” training videos into searchable, structured, and revenue-generating learning experiences. The AI-powered lesson-plan builder delivers conversational content generation directly inside their Learning Management System (LMS), helping learners find what they need quickly and saving the editorial team hours per week.
Industry
Use Case
AI-driven content discoverability and conversational lesson-plan generation
Solution implemented
- Built a conversational lesson-plan generator powered by AWS Bedrock and Claude v3 Sonnet, fine-tuned with petroleum engineering glossaries
- Ingested, transcribed, and chunked 820+ hours of video content using serverless AWS Lambda pipelines
- Developed hybrid search capabilities via Amazon OpenSearch with both vector and keyword indexing
- Deployed a lightweight React interface that integrates directly into Saga’s LMS for seamless user interaction
- Containerized architecture with ECS Fargate, FastAPI microservices, and DynamoDB for session caching and <2s median latency
- Provisioned all infrastructure using Terraform for scalable, replicable deployment
Outcomes expected
- 4x increase in average watch time per user within 60 days of rollout
- 93% topic match accuracy with lesson outlines generated in under 8 seconds
- 70% reduction in monthly AWS infrastructure costs vs EC2-based prototype
- 35% drop in inference-related costs due to optimized architecture and fine-tuning
- +35pt increase in NPS for content discoverability (from 27 to 62)
- Improved editorial productivity with automated lesson plans and source citations
Company Snapshot
Saga Wisdom is an on-demand learning platform that equips oil and gas professionals with expert-led, practical courses.
Location
Calgary, Alberta, Canada
Saga Wisdom’s AI Solution Makes 850+ Hours of Content Instantly Accessible
Saga Wisdom helps oil and gas professionals stay sharp with short, expert-led courses. But its core asset, over 850 hours of video, was largely inaccessible. With no transcripts or metadata, fewer than 5% of videos were being viewed each month. Editors had to manually scan footage to build courses, which slowed innovation and new content creation.
The goal was to unlock that video archive, make every moment discoverable, and turn raw footage into structured, searchable learning products. Saga partnered with New Math Data to bring that vision to life through a conversational, AI-powered lesson-plan builder, integrated directly into their Learning Management System (LMS).
The platform transcribes and segments videos automatically, indexes content for hybrid search, and uses a fine-tuned LLM to generate publication-ready outlines in seconds. The result: better discoverability, faster course production, and a massive leap in learner engagement.
Problem
Saga Wisdom had built an extraordinary library of expert video content, but much of its value remained locked away. Learners often rewatched familiar material instead of being able to go straight to the most relevant insights for their needs. Without topic-level search, countless microlearning opportunities went undiscovered, limiting engagement with the broader collection. Editors spent hours manually reviewing footage to structure courses, time that could have been invested in curating and amplifying the strongest material. The upside was clear: by making the content more discoverable and accessible, Saga Wisdom could have transformed underutilized assets into a dynamic growth driver, delivering more value to learners while increasing the return on its existing investment.
Saga needed a scalable, production-ready solution that could:
- Automatically transcribe and process hours of video footage
- Allow semantic and keyword-based search across all content
- Generate structured, editable lesson plans on demand
- Fit into their existing LMS infrastructure
- Be fast, low-latency, and cost-efficient to operate
Solution
Unlocking Learning Value from Unmonetizable Video Data
New Math Data designed and implemented a custom AI-powered solution to make every second of Saga’s training content searchable and structured. The system integrates directly with their LMS and delivers conversational access to curated course outlines.
The solution combines foundational LLM technology, video intelligence pipelines, hybrid search, and an intuitive front-end experience.
Conversational Lesson Planning with Bedrock
A FastAPI microservice calls Claude v3 Sonnet via AWS Bedrock, which has been fine-tuned with Saga’s petroleum engineering glossaries. The model generates publication-ready lesson plans, including titles, learning outcomes, and time-stamped references to the source video.
Video Processing at Scale
A serverless pipeline ingests hours of content using AWS Lambda, automatically producing transcripts and segmenting them into ~350-token chunks for optimal retrieval.
High-Precision Hybrid Search Index
By storing both embeddings and traditional text indices in Amazon OpenSearch, Saga’s users can instantly locate precise content matches using natural language queries or keywords.
Production-Ready Deployment
The full stack runs in containers on AWS ECS Fargate. DynamoDB caches session state for sub-2-second median response times. Terraform scripts provision everything from scratch, including networking, permissions, and monitoring.
Over a focused six-week engagement, New Math Data delivered an end-to-end, cloud-native stack that unlocks Saga Wisdom’s video archive and drops straight into its Learning Management System.
Related Case Studies
Vertically Integrated Utility Company
Migration of data systems and applications to AWS.
Ready to Transform Your Business?
See how New Math Data can transform your business with AWS-powered innovation.