Sold by: Klika Tech
Fixed-scope professional services engagement that adds adaptive on-device AI to deployed microcontroller-based IoT products using EdgeTNN, Amazon SageMaker AI, and AWS IoT Core.
Overview
Image
On-Device Adaptive AI Engagement Workflow
Image
On-Device Adaptive AI Reference Architecture
On-Device Adaptive AI for IoT Fleets
Manufacturers and operators of connected devices know that individual machines, sensors, meters, and environments rarely behave identically. Operating signals change as equipment wears, sensors age, and conditions evolve. A model trained only on centralized data may perform inconsistently across individual units -- and transmitting continuous raw sensor data to the cloud is often impractical due to bandwidth, power, cost, or privacy constraints.
Klika Tech solves this by bringing controlled model adaptation directly to compatible microcontrollers, keeping your existing hardware in service when the assessment confirms sufficient processing capacity, memory, and firmware headroom.
Built by a proven IoT and edge-AI team: Klika Tech is an AWS Premier Tier Services Partner holding the AWS IoT Competency and AWS IoT Core Service Delivery designation. The team is ISO 9001:2015 certified for quality management and ISO/IEC 27001:2022 certified for information-security management covering hardware, firmware, IoT, and cloud-native development.
How the Engagement Works
1. Data Audit and Model Training Klika Tech audits your available data or conducts a capture campaign using representative devices and their existing sensors. The team prepares the dataset, then trains and validates an initial backbone model in Amazon SageMaker AI against customer-approved acceptance criteria. The accepted model is optimized for the target microcontroller and packaged for secure deployment.
2. On-Device Learning with EdgeTNN Klika Tech integrates EdgeTNN, its proprietary C++ on-device learning library, into your firmware. EdgeTNN uses a frozen model backbone with a small trainable head that learns the normal operating behavior of each unit and adapts after legitimate changes such as maintenance or recalibration. Inference and fine-tuning run locally -- raw operational data stays on the device while only scores, events, and compact model-weight updates are transmitted to AWS.
3. AWS Reference Backend Deployment A complete AWS backend is deployed in your account using AWS Cloud Development Kit:
- AWS IoT Core or AWS IoT Core for LoRaWAN for device connectivity
- AWS IoT Device Management and AWS IoT Jobs for fleet indexing and OTA delivery of firmware and model packages
- Amazon Timestream for InfluxDB for score and event storage via AWS IoT rules
- Amazon Managed Grafana for operational dashboards
- Amazon EventBridge and Amazon SNS for anomaly alerts
- Amazon S3 for model archives and device-specific weight retention enabling auditability and rollback
- Amazon CloudWatch and AWS IAM for monitoring and access control
Built-In Safety and Governance
Model adaptation includes safeguards designed to limit unintended behavior:
- Each newly trained head is tested on-device against known-good reference data before activation
- Updates that fail validation are automatically discarded
- The original factory model remains available for rollback at all times
- Only the smaller model head is allowed to change -- the backbone stays frozen
- The system produces scores and alerts; maintenance, safety, and remediation decisions remain with your authorized personnel
Proven Edge-AI and MCU Expertise
Klika Tech brings deep experience in TinyML and edge machine learning, including MCU-based inference, local model adaptation, concept-drift management, and OTA model delivery. Demonstrated work includes:
- EdgeTNN integration with Infineon sensors and Amazon FreeRTOS for local industrial HVAC anomaly detection, with Amazon SageMaker AI for model training and optimization
- Optimized neural network deployment to STM32WBA52 and STM32WL55 devices, processing raw sensor data locally and sending only anomaly events to AWS
- Industrial anomaly detection for vibration, temperature, current, and acoustic signals in collaboration with STMicroelectronics, Infineon, and AWS
- Automated MLOps framework using Amazon SageMaker AI, AWS CodeBuild, hardware-specific optimization, and OTA deployment to resource-constrained devices
- Fleet-scale OTA delivery validated across more than 10,000 connected vehicles
- FreeRTOS porting across Nordic, STM32, ESP32, PIC32, Kinetis, SiFive, Intel, and Renesas platforms, including BLE, secure sockets, bootloaders, and OTA functionality
- Historical contributions to the FreeRTOS mainline for BLE, OTA, and bootloader support on the nRF52840
Klika Tech is also a LoRa Alliance member, supporting customers using the AWS IoT Core for LoRaWAN connectivity path.
Highlights
- Modernize Deployed Devices Without Hardware Replacement - Add adaptive edge AI to compatible microcontroller-based products already operating in the field using Klika Tech's proprietary EdgeTNN library. Klika Tech has demonstrated EdgeTNN integration on Infineon and STMicroelectronics MCU platforms, including STM32WBA52 and STM32WL55, processing raw sensor data locally and sending only anomaly events to AWS.
- Improve Accuracy with Private, On-Device Learning - Enable each device to adapt to its unique operating conditions and changing signals while keeping raw operational data local. Only scores, events, and compact model updates are transmitted. Klika Tech has implemented edge anomaly detection across vibration, temperature, current, and acoustic signals for industrial equipment, with Amazon SageMaker AI supporting model training and optimization.
- Deploy a Complete Device-to-AWS Foundation - Receive integrated EdgeTNN firmware, a validated initial model trained in Amazon SageMaker AI, secure OTA workflows, telemetry dashboards, alerts, and customer-owned AWS CDK infrastructure. The reference architecture uses AWS IoT Core for device connectivity and includes model rollback safeguards.
Details
Sold by
Delivery method
Deployed on AWS
New
Introducing multi-product solutions
You can now purchase comprehensive solutions tailored to use cases and industries.
Pricing
Custom pricing options
Pricing is based on your specific requirements and eligibility. To get a custom quote for your needs, request a private offer.
Legal
Content disclaimer
Vendors are responsible for their product descriptions and other product content. AWS does not warrant that vendors' product descriptions or other product content are accurate, complete, reliable, current, or error-free.
Support
Vendor support
Klika Tech provides professional services support throughout the full engagement lifecycle: device compatibility assessment, data review or collection, model development, firmware integration, AWS deployment, testing, OTA validation, and operational handoff.
Engagement Team Customers are supported by a dedicated team that may include embedded engineers, machine learning specialists, AWS cloud engineers, solution architects, and technical project leadership. Klika Tech is an AWS Premier Tier Services Partner with AWS IoT Competency and AWS IoT Core Service Delivery designation, and holds ISO 9001:2015 (quality management) and ISO/IEC 27001:2022 (information security) certifications.
Support Activities Support includes technical workshops, model and device acceptance criteria, integration troubleshooting, deployment guidance, validation of on-device learning and rollback safeguards, knowledge transfer, and documentation. Engagement scope, schedule, communication channels, and response expectations are established in the statement of work based on device hardware, firmware, data availability, connectivity, and model requirements.
Post-Engagement Support Post-launch model tuning, fleet expansion, monitoring enhancements, and ongoing engineering support are available through a separately scoped engagement.
Contact Information
Support is provided by the assigned Klika Tech engagement team throughout project delivery, validation, deployment, and transition activities.