Browse categories
Explore
Fiverr Pro
English
$
USD
I will develop custom tinyml models for microcontrollers and single board computers
Saran Khaliq
About this gig
Bring intelligence to the edge! I develop and optimize high-performance AI models for resource-constrained hardware like ESP32, ESP32-S3, Raspberry Pi, and STM32. If you need to run complex Machine Learning on microcontrollers with minimal RAM and Flash, I am here to help.
My Expertise Includes:
- Custom AI Development: Optimized architectures (CNN, RNN, TinyYOLO, MobileNet) for Edge devices.
- Model Optimization: Advanced Post-Training Quantization (INT8/Float16) and Pruning to reduce size and power usage.
- Deployment: Production-ready TensorFlow Lite Micro (TFLite) and C++ code for Arduino/ESP-IDF.
- Computer Vision: Image classification and object detection for ESP32-Cam.
- Signal Processing: AI for IMU sensors, keyword spotting, and anomaly detection.
Supported Tech:
- Frameworks: TensorFlow Lite, PyTorch, Edge Impulse, Keras.
- Hardware: ESP32 series, Raspberry Pi 4/5/Pico, Arduino, ARM Cortex-M.
I bridge the gap between heavy Data Science and embedded silicon.
Contact me before ordering to discuss your hardware constraints
Get to know Saran Khaliq
Saran Khaliq
Building smarter solutions with Computer visions, AI and Robotics
- FromPakistan
- Member sinceApr 2021
- Avg. response time1 hour
- Last delivery1 week
Languages
Urdu, Hindi, English
👋 I’m a Machine Learning & AI Developer and Founder of BrainSwarm Robotics. I specialize in Deep Learning, Reinforcement Learning, ROS (Robot Operating System), NVIDIA Jetsons, NVIDIA DeepStream, and real-time edge computing for robotics and automation. I develop intelligent systems using Python, PyTorch, TensorFlow, and OpenCV, with expertise in computer vision, object detection, segmentation, and real-time video analytics. My goal is to deliver high-quality AI solutions for robotics, IoT, automation, and smart vision systems with efficiency and precision.
Other AI Development Services I Offer
FAQ
1. Can any AI model run on an ESP32 or ESP32-S3?
Not directly. Standard models are too large for microcontroller RAM. I specialize in Quantization (converting 32-bit to 8-bit) and Pruning to compress models so they fit on the ESP32’s limited memory without losing significant accuracy.
2. Which frameworks do you use for Embedded AI?
2. Which frameworks do you use for Embedded AI? I primarily work with TensorFlow Lite Micro (TFLite), Edge Impulse, ESPDL, ESP-PPQ (Quantization Tools) and PyTorch. For deployment, I provide C++ or Arduino-compatible code ready to be flashed onto your device via ESP-IDF or Arduino IDE.
3. Do I need to provide the dataset for training?
Ideally, yes. For custom tasks like specific object detection or unique sensor signals, a high-quality dataset is required. However, if you don't have one, I can assist in sourcing open-source data or advising you on how to collect it from your hardware.
4. What is the difference between Cloud AI and Edge AI (TinyML)?
Yes. If your current model is lagging or crashing your ESP32/Pi, I can apply optimization techniques like layer fusion and INT8 quantization to increase inference speed (FPS) and reduce the memory footprint.
5. Can you optimize my existing model to run faster?
Yes. If your current model is lagging or crashing your ESP32/Pi, I can apply optimization techniques like layer fusion and INT8 quantization to increase inference speed (FPS) and reduce the memory footprint.
