Senior Technical Architect (IN)
HCLTech
HCLTech
Spearhead the on-device AI execution stack for ARM-based SoCs, focusing on model conversion, quantization, and runtime integration. This role is crucial for deploying vision, LLM, and VLM workloads efficiently within strict edge power and memory constraints.
This is a senior position requiring extensive experience in optimizing AI models for embedded systems. You will be instrumental in pushing the boundaries of on-device AI capabilities.
Deploy and optimize neural network models across NPU, GPU, and CPU backends utilizing embedded inference runtimes. You will own the model conversion pipeline, including graph capture, operator mapping, quantization (PTQ/QAT), calibration, and accuracy validation.
Diagnose and resolve issues related to unsupported operators and graph partitioning. Enable on-device LLM and VLM workloads through weight quantization and performance tuning. Benchmark and optimize inference latency, throughput, and energy efficiency, publishing reproducible performance data.
Integrate inference into real-time media and robotics pipelines with efficient tensor sharing. Develop robust tooling for model deployment and regression testing to ensure accuracy and performance consistency. Advise product and data-science teams on optimal model architectures for edge accelerators.
A minimum of 8-10 years in software engineering, with at least 3 years dedicated to on-device or edge AI deployment, is essential. Proficiency in C++ and Python is required, alongside hands-on experience with embedded inference runtimes such as LiteRT/TFLite, ONNX Runtime, ExecuTorch, or vendor NPU SDKs.
Practical expertise in quantization techniques (INT8/INT4, per-channel schemes, calibration, accuracy recovery) and model format conversions (PyTorch/TensorFlow to deployable graphs) is necessary. You should be adept at profiling on heterogeneous SoCs and understanding memory bandwidth constraints. A solid grasp of CNN, transformer, and modern vision/language model architectures is expected.
HCLTech
IT Consulting