Bangalore, Karnataka
Job Summary

We are building a team to take research-grade computer vision algorithms and turn them into shippable, production-quality embedded software for AI-enabled microcontroller platforms — covering capabilities such as object/person detection, gesture recognition, and presence sensing. Reference algorithm models developed by an upstream research group must be converted, optimized, and adapted into robust embedded C, partitioned across DSP and NPU/AI-accelerator hardware, and tuned per target platform and market.

Key Responsibilities
Convert computer vision algorithm reference implementations (e.g., MATLAB/Python prototype models) into optimized, production-quality embedded C.
Perform platform-specific adaptation: partition vision pipelines across a heterogeneous compute architecture where heavy pre/post-processing is offloaded to a DSP and inference runs on a dedicated NPU/AI accelerator, optimizing for low power and cost.
Perform market-specific adaptation: tune and validate detection/recognition behavior for different deployment environments, lighting conditions, and customer/market requirements.
Design and implement embedded computer vision pipelines (image capture, pre-processing, inference, post-processing) in C/C++ on MCU-class hardware.
Integrate camera input pipelines (e.g., via USB or standard camera-interface protocols) and manage frame buffering, format conversion, and pre-processing for downstream inference.
Port, optimize, and validate quantized neural network models (object detection, classification, gesture/pose recognition) for on-device NPU/DSP-accelerated inference.
Optimize vision inference pipelines for latency, frame rate, memory footprint, and power consumption on constrained edge hardware.
Work with graphics/display subsystems (e.g., GUI/HMI frameworks) where vision results feed into on-device user interfaces.




Collaborate closely with the upstream algorithm research team to validate that productized C implementations preserve reference accuracy and behavior.
Debug and profile using hardware debuggers, and support board bring-up and system integration testing.
Document pipeline architecture, model integration steps, and performance benchmarks.
Skill Requirements
Robust embedded C/C++ development experience on Arm Cortex-M (or similar) microcontrollers.
Solid understanding of computer vision fundamentals: image pre-processing, object detection/classification concepts, and common model architectures used for embedded vision.
Experience deploying and optimizing on-device ML inference — familiarity with TensorFlow Lite Micro, CMSIS-NN, or comparable embedded inference runtimes/compilers, including model quantization.
Experience with camera/video data pipelines (frame capture, buffering, color format conversion) in an embedded or resource-constrained context.
Working knowledge of RTOS-based embedded systems and low-power design patterns.
Comfortable with hardware bring-up and debugging (JTAG/SWD, logic analyzers).
Development of unit-tests and system-tests
Scripting (Python) for automated testing
ARM, DSP programming experience
Version control (Git) and embedded build/test tooling.
Other Requirements
Familiarity with GUI/embedded graphics frameworks (e.g., LVGL or similar) for vision-driven HMI applications.
Experience with MIPI camera/display interfaces (CSI/DSI) or equivalent embedded imaging standards.
Exposure to model compression, pruning, or hardware-aware neural architecture optimization.
Prior experience in smart-home cameras, robotics, kiosks, or industrial vision products.
Python for model evaluation, dataset tooling, or test automation.

Bachelor’s or Master’s degree in Electronics, Computer Engineering, Computer Science, or related field.

#body.unify div.unify-button-container .unify-apply-now: focus, #body.unify div.unify-button-container .unify-apply-#body.unify div.unify-button-container .unify-apply-now: focus, #body.unify div.unify-button-container .unify-apply-

📌 Senior Technical Architect (India)
🏢 HCLTech
📍 India