Data is never the problem. Accessing it is.
In design-driven industries like construction and engineering, fast access to project data can mean the difference between smooth coordination and costly delays. But when key insights are buried in spreadsheets, custom reports, or inaccessible systems, even well-run teams slow down.
A multidisciplinary MEP and fire protection firm operating across commercial, institutional, and healthcare projects in North America, our client manages large volumes of operational and QA data spanning project schedules, design milestones, and contractor workflows. However, non-technical teams struggled to access or analyze this data independently. Reporting cycles were slow, dashboards took days to update, and inter-team collaboration relied heavily on manual coordination.
Coditas built a secure, AWS-native analytics platform that lets users ask questions in plain English and get instant, visual answers. No SQL. No waiting. No ticketing system. Just a faster, self-serve way to turn internal data into real-time, decision-ready insight, built entirely on AWS.
The platform turns natural language into real-time dashboards. At its core is Claude 3.5 Sonnet, accessed via Amazon Bedrock, which interprets user questions and converts them into SQL. The platform then queries Amazon RDS and renders the results as interactive reports without engineering involvement.
- Amazon Bedrock with Claude 3.5 Sonnet interprets user queries and returns optimized SQL
- Amazon RDS executes those queries on DHC’s internal data, returning structured results
- ECS with Docker Compose handles backend services in a modular, containerized setup
- AWS Lambda manages inference routing, keeping latency low without maintaining servers
- Amazon S3 stores user files and downloadable reports
- CloudWatch and Systems Manager provide observability, diagnostics, and logging
- IAM + private subnets + TLS encryption lock down access and protect PHI/PII end to end
Once the platform went live, things moved fast. And for the first time, so did the data.
-
75% faster dashboard creation
→ Reduced reporting cycles from ~2 days to under 30 minutes -
50% improvement in decision-making speed
→ Teams accessed live project data during reviews, not after -
100% adoption within 2 weeks
→ Used daily by project managers, QA leads, and admin teams across departments -
20+ engineering hours saved per week
→ Shifted support bandwidth from manual report requests to core platform enhancements -
40% reduction in follow-up delays across inter-department escalations
→ Real-time dashboards replaced versioned spreadsheets and email bottlenecks -
Compliance passed on first review
→ All data encrypted, access managed through IAM, and no PII/PHI exposed to LLMs
We cleanly embed GenAI while wrapping it in AWS best practices.
- Prompt engineering over fine-tuning: We scoped and versioned templates tailored to real reporting use cases
- Secure-by-design architecture: VPC segmentation, IAM roles, and TLS enforced across all layers
- Modular, containerized backend: ECS and Docker Compose let us ship fast and scale clean
- No model management required: Using Bedrock kept LLM ops minimal and production-ready
- Tight feedback loop: Prompt logic was continuously improved based on real user behavior
The result? A real product built for real users, with fast results, secure infra, and a roadmap that includes multilingual queries, root cause analysis, and trend detection.
Need help with your Business?
Don’t worry, we’ve got your back.
