Embedded AI Engineer

FONTES SINGAPORE PRIVATE LIMITED; Singapore; On-site; SGD 100,000 - 156,000; Full time

Job summary

FONTES SINGAPORE PRIVATE LIMITED is seeking an Embedded AI Engineer to build, optimize, and deploy neural networks on ultra-low-power, memory-constrained SoCs and AI accelerators. You will convert trained models into efficient on-device inference pipelines for battery-powered edge devices across vision, audio, activity, and time-series domains.

You will deploy AI models to proprietary hardware, optimize for latency and power, perform INT8 quantization, and collaborate with hardware and firmware

Qualifications

• Strong grounding in machine learning fundamentals.
• 0
• Experience deploying AI on embedded or edge hardware.
• 0
• Proficiency in C/C++ (embedded) and Python.
• 0
• Hands-on experience with model quantization (PTQ or QAT).
• 0
• Familiarity with memory-constrained system design.
• 0
• Ability to optimize for latency, memory footprint, and power draw.
• 0

Responsibilities

• Deploy AI models onto resource-constrained, proprietary SoC designs.
• Optimize models across memory, power, and latency dimensions.
• Execute model quantization at INT8 or lower precision.
• Map neural networks to microcontrollers, DSPs, and AI accelerators.
• Build inference pipelines in bare metal or Zephyr OS (advantageous).
• Validate embedded inference outputs against Python reference models.
• Diagnose and resolve numerical mismatches between floating-point and quantized inference.
• Tune memory layout, buffer reuse strategies, and tensor ordering.
• Work closely with hardware and firmware engineering teams.

Skills

ML fundamentals

Embedded AI deployment

C/C++ (embedded)

Python

Quantization PTQ/QAT

Memory-constrained design

Latency/Memory/Power optimization

Tools

TensorFlow Lite Micro

TVM

CMSIS-NN

ONNX Runtime

Job description

About the Role

We are looking for an Embedded AI Engineer to build, optimize, and deploy neural networks on ultra-low-power, memory-constrained SoCs and AI accelerators. This role sits at the intersection of AI, embedded systems, and silicon - converting trained models into highly efficient on-device inference pipelines for battery-powered and always-on edge applications across vision, audio, activity, and time-series domains.

Responsibilities

• Deploy AI models onto resource-constrained, proprietary SoC designs
• Optimize models across memory, power, and latency dimensions
• Execute model quantization at INT8 or lower precision
• Map neural networks to microcontrollers, DSPs, and AI accelerators
• Build inference pipelines in bare metal or Zephyr OS (advantageous)
• Validate embedded inference outputs against Python reference models
• Diagnose and resolve numerical mismatches between floating-point and quantized inference
• Tune memory layout, buffer reuse strategies, and tensor ordering
• Work closely with hardware and firmware engineering teams

Requirements

• Solid grounding in machine learning fundamentals
• Practical experience deploying AI on embedded or edge hardware
• Strong proficiency in C/C++ (embedded context) and Python
• Hands-on experience with quantization techniques - PTQ or QAT
• Familiarity with memory-constrained system design
• Demonstrated ability to optimize for latency, memory footprint, and power draw

Bonus Skills

• Exposure to TensorFlow Lite Micro, TVM, CMSIS-NN, or ONNX Runtime
• Knowledge of RISC-V architecture or custom AI accelerator design
• Experience with edge sensors - audio, IMU, vision, or wearables
• Background in power profiling and power optimization workflows
• Agile development exposure

Why Join

• Work directly on production silicon and real edge-AI systems
• Build AI designed to run for months or years on battery power
• Direct collaboration with chip architects and systems engineers
• End-to-end ownership from model training to silicon deployment
• High-impact position within a fast-moving deep-tech environment