AI, Machine learning, Robotics, Embedded systems, edge AI, Full Stack Web Development & automotive test engineering.
From bare-metal firmware and HIL/SIL rigs to on-device ML,one partner across the whole path, from silicon to cloud.
- Embedded
- Machine Learning
- Firmware
- RTOS
- CAN / TSN
- HIL / SIL
- Edge AI
- Cloud
Services
One partner across the whole path, from silicon to cloud.
Embedded & Firmware
Board bring-up to production firmware on constrained hardware.
- C/C++
- Python
- Embedded Linux
- BSP / Drivers
- FreeRTOS
- Bare-metal
- STM32
- ESP32
- Texas Instruments
- Arduino
- Raspberry Pi
- Raspberry Pico
- Odroid
Automotive Test
HIL/SIL rigs that validate ECUs before they ship.
- HIL/SIL
- NI VeriStand
- TwinCAT
- Aliaro SIL
- CANoe
- Structured Text
- Python
Edge AI / TinyML
Quantized computer vision that runs on-device, not in the cloud.
- PyTorch
- TFLite Micro
- YOLOv8
- Quantization
- On-device CV
Vehicle Networking
Reliable, real-time communication between ECUs and sensors.
- CAN / CAN FD
- LIN
- SPI / I2C / UART
- TSN
- EtherCAT / Modbus
- Automotive Ethernet
- TCP/IP
IoT & Connectivity
Connected devices with secure telemetry from edge to cloud.
- MQTT
- BLE 5.2 / Wi-Fi
- LoRa
- 4G
- TCP/IP
- AWS IoT
- GCP
AI/ML & Computer Vision
Deep-learning models and computer vision, from research to deployment.
- Python
- Deep Learning
- YOLO
- PyTorch
- TensorFlow
- OpenCV
Web & Full-Stack Development
Responsive web apps and APIs — the full front-to-back software layer.
- React / Next.js
- C# / GUIs
- Python
- REST APIs
- HTML / CSS / JS
- SQL
- GCP
CI/CD & Test Automation
Automated pipelines and regression testing that keep releases reliable.
- CI/CD Pipelines
- Git
- Regression Testing
- Test Automation
Data Annotation & Labeling
High-quality labeled datasets for training and validating ML models.
- Image Labeling
- Bounding Boxes
- Segmentation
- Datasets
- QA
Case studies
Anonymized proof, problem, approach, and the result that mattered.
- Edge AI
On-device detection with instant alerting
- Problem
- A client needed real-time object detection on a low-cost, battery-friendly camera — without streaming video to the cloud.
- Approach
- Deployed a quantized YOLOv8n model on an ESP32-CAM, publishing detections over MQTT for immediate alerting.
- Result
- Sub-second on-device alerts with no cloud video costs and a fraction of the bandwidth.
- Computer Vision
Detection tuned for Nordic winter conditions
- Problem
- Vehicles had to be detected reliably in heavy snow and low light, where off-the-shelf models degrade sharply.
- Approach
- Benchmarked and tuned detection models against a Nordic winter dataset, measuring the accuracy-vs-latency trade-off.
- Result
- Identified the best accuracy/latency configuration for deployment on constrained hardware.
- Vehicle Networking
Deterministic real-time pipeline over TSN + 4G
- Problem
- A system needed low-latency, deterministic data transport across in-vehicle and remote links.
- Approach
- Integrated Time-Sensitive Networking (TSN) with a 4G uplink for synchronized, real-time telemetry.
- Result
- Deterministic in-vehicle timing with reliable remote streaming to the cloud.
- Predictive AI
Predictive maintenance for HIL test systems
- Problem
- Unexpected hardware degradation in HIL test rigs caused unplanned downtime during ADAS validation.
- Approach
- Built predictive-maintenance models on test-system telemetry to flag early degradation, integrated into existing HIL workflows.
- Result
- Earlier fault detection and reduced unplanned downtime for the test infrastructure.
- Sensor Fusion
Indoor localization with BLE and sensor fusion
- Problem
- Indoor positioning needed higher accuracy than signal-strength (RSSI) alone could deliver.
- Approach
- Fused BLE 5.2 RSSI fingerprinting with Angle-of-Arrival data through an ANN model.
- Result
- Improved localization accuracy across varying network and environmental conditions.
- Automotive Test
End-to-end HIL rig for safety-critical bus systems
- Problem
- Safety-critical vehicle systems needed automated, repeatable verification before deployment.
- Approach
- Designed and commissioned HIL rigs end-to-end — from physical build to NI VeriStand test environments and custom control GUIs.
- Result
- Automated validation aligned with automotive safety and quality standards.
- TinyML
Always-on TinyML sensing on a microcontroller
- Problem
- A product needed always-on recognition on a battery-powered microcontroller, with no cloud connection.
- Approach
- Trained and quantized a compact neural network, deploying it with TensorFlow Lite Micro on an ESP32 / STM32-class device.
- Result
- Real-time, offline inference within tight memory and power budgets.
- GUI & Tooling
Desktop GUIs for hardware control and test
- Problem
- Engineers needed a simple desktop interface to configure, control, and monitor test hardware — without touching low-level scripts.
- Approach
- Built control GUIs in C# and Python that drive the hardware over serial/CAN and visualize live data in real time.
- Result
- Faster, repeatable test setup with live control and monitoring from a single interface.
- Firmware
Portable bare-metal firmware across MCU families
- Problem
- A product line needed consistent low-level drivers across several microcontroller families.
- Approach
- Wrote portable C/C++ bare-metal firmware and BSP/drivers (SPI, I2C, UART, CAN) targeting STM32, ESP32, Texas Instruments, Arduino, Raspberry Pi/Pico and Odroid.
- Result
- A reusable driver layer that brings new boards up quickly with minimal rework.
- RTOS Firmware
Real-time firmware on FreeRTOS across microcontrollers
- Problem
- Time-critical tasks needed deterministic scheduling on constrained hardware.
- Approach
- Developed C/C++ real-time applications on FreeRTOS, deployed across STM32, ESP32, Raspberry Pico and Arduino-class devices.
- Result
- Deterministic task timing and stable long-run operation on low-power hardware.
Tech stack
The tools, platforms, and frameworks behind the work.
Languages
- C
- C++
- Python
- C#
- Structured Text (IEC 61131-3)
- VHDL
- JavaScript
- SQL
Embedded & RTOS
- FreeRTOS
- MicroC/OS-II
- Rubus RTOS
- Bare-metal
- BSP / Drivers
- Embedded Linux
- PLCs
Hardware
- STM32
- ESP32
- Texas Instruments
- Arduino
- Raspberry Pi
- Raspberry Pico
- Odroid
Protocols & Networking
- CAN / CAN FD
- SPI / I2C / UART
- TSN
- BLE 5.2
- 4G
- EtherCAT
- Modbus
- Ethernet / TCP/IP
Test & Tooling
- HIL / SIL
- NI VeriStand
- TwinCAT
- Aliaro SIL
- Test Automation
- CI/CD
- Git
AI / ML
- PyTorch
- TensorFlow
- Keras
- OpenCV
- TFLite Micro
- YOLO
- Edge AI / TinyML
Cloud & Web
- Google Cloud
- React / Next.js
- REST APIs
- HTML / CSS / JS
About
zaramTech AB is a boutique engineering consultancy spanning the full path from silicon to cloud — embedded firmware, HIL/SIL automotive test, on-device edge AI, and full-stack software.
Behind it is 10+ years designing, integrating, and validating embedded and automotive software. At Volvo Buses, that means building Hardware-in-the-Loop rigs end-to-end and the automated test infrastructure — NI VeriStand, TwinCAT, and Aliaro SIL — that validates safety-critical vehicle systems.
That deep firmware, RTOS, and vehicle-networking background is paired with applied AI/ML research — including a quantized YOLOv8n detector deployed on an ESP32-CAM via TFLite Micro for real-time, offline, on-device detection.
Credibility
- 10+ years in embedded & automotive
- Volvo Buses — HIL/SIL test engineering
- MSc Computer Science · M.Eng Applied AI
- Swedish Institute & Google Africa scholar
Let's talk about your project
Embedded development, AI, Machine Learning, Robotics, HIL/SIL, edge AI, or full-stack — tell me what you're building and I'll get back to you.