Honeywell HVAC units streaming live telemetry to AWS IoT Core
Database servers to manage, with S3, Glue and Athena as the data lake
Of the environment defined in Terraform, ready to extend to new buildings
SparkAI
Dover, DE
AI-Powered Building Automation / Smart Energy Management
AWS IoT Core, Amazon S3, AWS Glue, Amazon Athena, Amazon QuickSight, Amazon API Gateway, AWS Lambda, Amazon CloudWatch, AWS IAM, AWS KMS, Terraform
SparkAI is building a platform that coordinates HVAC, lighting, and EV charging systems in commercial buildings for intelligent energy management, systems that today run independently with no shared coordination. Ahead of a pilot deployment with the National Renewable Energy Laboratory, Avahi built the AWS data foundation connecting SparkAI’s on-site hardware to a cloud data lake and analytics layer, giving SparkAI live telemetry, historical data, and query access to build on for the platform’s next phase.
SparkAI is developing an AI-powered building automation platform for commercial buildings, with partnerships in place with the National Renewable Energy Laboratory for pilot deployment and Nvidia for edge computing hardware.
Commercial building systems like HVAC, lighting, and EV chargers are typically installed independently and never talk to each other. When utilities want a building to reduce load, the usual path is a text message to a facility manager, who then manually adjusts equipment, with no automated execution, no system of record, and no way to prove compliance after the fact. SparkAI set out to replace that manual process with an AI-coordinated system, but first needed a real data pipeline connecting building equipment to the cloud, since the platform was being built from the ground up with no operational system yet in place.
SparkAI needed a data pipeline that could ingest live telemetry from building equipment on a tight polling interval alongside a large historical dataset for future model training, and query it back out through fast, ad hoc analytics. AWS IoT Core provided a secure, certificate-authenticated ingestion path from SparkAI’s on-site hardware, while Amazon S3, AWS Glue, and Amazon Athena gave the platform a queryable data lake without standing up and managing database infrastructure. The entire environment was provisioned as infrastructure as code with Terraform, so SparkAI can extend it as the platform scales to more buildings.
As a premier-tier AWS partner with experience standing up IoT data pipelines from scratch, Avahi could take SparkAI from an early-stage, ground-up build to a working data foundation, integrating directly with SparkAI’s own hardware rather than a simulated environment. That hands-on device integration work, combined with a data lake architecture built to scale from a single pilot site toward SparkAI’s much larger multi-facility vision, gave SparkAI a foundation it can carry directly into its next build phase.
Avahi connected SparkAI’s on-site Control Box to its four Honeywell HVAC units using the BACnet protocol, collecting device telemetry on a frequent polling interval with error handling and connection management built in. The Control Box runtime aggregates that telemetry and syncs it to AWS over an encrypted MQTT connection into AWS IoT Core, with a safety rules engine in place to enforce configured temperature thresholds.
On the cloud side, Avahi built a serverless data lake on Amazon S3, with raw and processed zones, lifecycle policies, and encryption, fed by both live device telemetry and a large historical building-energy dataset from SparkAI’s NREL partnership. AWS Glue handles schema cataloging and the ETL jobs that clean and normalize incoming data, and Amazon Athena provides fast, ad hoc SQL access across the resulting data lake. A REST API layer, backed by Amazon API Gateway and AWS Lambda, exposes current and historical device readings, and a basic Amazon QuickSight dashboard gives SparkAI’s team a live view into telemetry across their HVAC units. The full environment, networking, IoT ingestion, data lake, and supporting security and monitoring, was provisioned as infrastructure as code so it can be extended as SparkAI adds buildings and devices.
The platform’s edge AI inference and closed-loop control execution, which will let the system act on this data automatically, are being built in a following phase, on top of the data foundation delivered here.
BACnet-based telemetry integration with SparkAI’s on-site Control Box and HVAC units
A Control Box runtime with telemetry aggregation, a configurable safety rules engine, and encrypted cloud sync to AWS IoT Core
A serverless AWS data lake (S3, Glue, Athena) combining live device telemetry with a large historical energy dataset
A REST API layer for current and historical device data, backed by API Gateway and Lambda
A basic analytics dashboard for HVAC telemetry, built on Amazon QuickSight
The full environment provisioned as infrastructure as code with Terraform, documented for SparkAI’s team to extend
SparkAI moved from an early-stage concept with no operational system to a real, working data pipeline connecting its on-site hardware to a cloud data lake it can query, analyze, and build on. With telemetry flowing, historical data loaded, and the infrastructure defined as code, SparkAI has the foundation it needs to move into the next phase of the platform: adding the edge AI inference and automated control that turns this data into action.
SparkAI
Dover, DE
AI-Powered Building Automation / Smart Energy Management
AWS IoT Core, Amazon S3, AWS Glue, Amazon Athena, Amazon QuickSight, Amazon API Gateway, AWS Lambda, Amazon CloudWatch, AWS IAM, AWS KMS, Terraform
SparkAI is building a platform that coordinates HVAC, lighting, and EV charging systems in commercial buildings for intelligent energy management, systems that today run independently with no shared coordination. Ahead of a pilot deployment with the National Renewable Energy Laboratory, Avahi built the AWS data foundation connecting SparkAI’s on-site hardware to a cloud data lake and analytics layer, giving SparkAI live telemetry, historical data, and query access to build on for the platform’s next phase.
SparkAI is developing an AI-powered building automation platform for commercial buildings, with partnerships in place with the National Renewable Energy Laboratory for pilot deployment and Nvidia for edge computing hardware.
Commercial building systems like HVAC, lighting, and EV chargers are typically installed independently and never talk to each other. When utilities want a building to reduce load, the usual path is a text message to a facility manager, who then manually adjusts equipment, with no automated execution, no system of record, and no way to prove compliance after the fact. SparkAI set out to replace that manual process with an AI-coordinated system, but first needed a real data pipeline connecting building equipment to the cloud, since the platform was being built from the ground up with no operational system yet in place.
SparkAI needed a data pipeline that could ingest live telemetry from building equipment on a tight polling interval alongside a large historical dataset for future model training, and query it back out through fast, ad hoc analytics. AWS IoT Core provided a secure, certificate-authenticated ingestion path from SparkAI’s on-site hardware, while Amazon S3, AWS Glue, and Amazon Athena gave the platform a queryable data lake without standing up and managing database infrastructure. The entire environment was provisioned as infrastructure as code with Terraform, so SparkAI can extend it as the platform scales to more buildings.
As a premier-tier AWS partner with experience standing up IoT data pipelines from scratch, Avahi could take SparkAI from an early-stage, ground-up build to a working data foundation, integrating directly with SparkAI’s own hardware rather than a simulated environment. That hands-on device integration work, combined with a data lake architecture built to scale from a single pilot site toward SparkAI’s much larger multi-facility vision, gave SparkAI a foundation it can carry directly into its next build phase.
Avahi connected SparkAI’s on-site Control Box to its four Honeywell HVAC units using the BACnet protocol, collecting device telemetry on a frequent polling interval with error handling and connection management built in. The Control Box runtime aggregates that telemetry and syncs it to AWS over an encrypted MQTT connection into AWS IoT Core, with a safety rules engine in place to enforce configured temperature thresholds.
On the cloud side, Avahi built a serverless data lake on Amazon S3, with raw and processed zones, lifecycle policies, and encryption, fed by both live device telemetry and a large historical building-energy dataset from SparkAI’s NREL partnership. AWS Glue handles schema cataloging and the ETL jobs that clean and normalize incoming data, and Amazon Athena provides fast, ad hoc SQL access across the resulting data lake. A REST API layer, backed by Amazon API Gateway and AWS Lambda, exposes current and historical device readings, and a basic Amazon QuickSight dashboard gives SparkAI’s team a live view into telemetry across their HVAC units. The full environment, networking, IoT ingestion, data lake, and supporting security and monitoring, was provisioned as infrastructure as code so it can be extended as SparkAI adds buildings and devices.
The platform’s edge AI inference and closed-loop control execution, which will let the system act on this data automatically, are being built in a following phase, on top of the data foundation delivered here.
BACnet-based telemetry integration with SparkAI’s on-site Control Box and HVAC units
A Control Box runtime with telemetry aggregation, a configurable safety rules engine, and encrypted cloud sync to AWS IoT Core
A serverless AWS data lake (S3, Glue, Athena) combining live device telemetry with a large historical energy dataset
A REST API layer for current and historical device data, backed by API Gateway and Lambda
A basic analytics dashboard for HVAC telemetry, built on Amazon QuickSight
The full environment provisioned as infrastructure as code with Terraform, documented for SparkAI’s team to extend
SparkAI moved from an early-stage concept with no operational system to a real, working data pipeline connecting its on-site hardware to a cloud data lake it can query, analyze, and build on. With telemetry flowing, historical data loaded, and the infrastructure defined as code, SparkAI has the foundation it needs to move into the next phase of the platform: adding the edge AI inference and automated control that turns this data into action.
Let’s explore your high-impact AI opportunities together in a complimentary session
