Software Engineer-Platform Device Evaluation Engineer (Seoul, South Korea)

Position Overview

MangoBoost builds full-stack AI infrastructure for next-generation hyperscale AI data centers. Our work covers AI networking, next-gen GPU systems and high-performance storage, along with the software ecosystem that ties them together. Within this stack, the DPU (Data Processing Unit) is an FPGA- or SoC-based PCIe card that offloads and accelerates network and storage processing from the CPU.

Our team develops the software that runs the Platform Device, a component shared across the DPU solutions deployed in AI infrastructure. This includes:

• Firmware that controls the card’s internal hardware
• Discovery and initialization of the PCIe card on Embedded Linux and on x86/Arm host servers
• Network interface drivers
• Management functions such as health monitoring, firmware updates and diagnostics

In this role, you will develop a common evaluation framework that spans every layer from firmware to CLI tools. It will let this software be evaluated against the same criteria on DPU products built on FPGAs and SoCs from multiple vendors. Building on it, you will automatically verify the functionality, compatibility, reliability and performance of each layer and of the integration between layers. You will also work with the development teams to trace defects to their root causes, so the card runs reliably under any HW/SW combination. You should be able to build environments ranging from hardware-free emulation to automated test labs with real cards attached. You should also be able to design test code and tools that can be reused as the product line keeps growing.

Responsibilities & Opportunities

• Evaluation strategy: Define per-layer evaluation strategies, test items and pass criteria for the platform software stack.
• Test frameworks: Develop unit, integration and system test frameworks for each layer. Build common libraries and tools for device control and result collection that can be reused across HW/SW versions.
• Hardware-free validation: Build emulation and mock-device environments (Mock, QEMU, etc.) to validate drivers, embedded Linux and firmware.
• HIL lab and CI pipeline: Build test labs with real cards attached. Automate CI regression testing that chains build, flashing, remote reset, testing and log collection, and automatically recover hung equipment.
• Functional validation: Cover device boot and discovery, driver installation and operation, network TX/RX, CLI and management functions, and firmware update and recovery.
• Compatibility evaluation: Run matrix testing across cross-layer version combinations, Linux distributions and kernel versions, and hardware configurations.
• Reliability and stress evaluation: Verify behavior and recovery under repeated resets, hot-plug, restarts and firmware updates, long-duration load, and fault injection.
• Performance evaluation: Measure throughput, latency and resource usage of the driver’s data/control paths and of management functions.
• Results management: Track evaluation results along with execution environments and version history. Build reports and visualizations that show functional and performance regressions against previous versions, and document evaluation methods and results.
• Defect analysis: Document reproduction conditions and narrow the root-cause scope using logs and execution flow. Work with the hardware, firmware and driver teams on root-cause analysis and verify that issues don’t recur.

Required Qualifications

• BS or higher in Computer Science, Electrical Engineering or a related field, or equivalent practical experience
• 3+ years of experience in system software development or test automation development, including industry and research projects
• Hands-on experience developing test automation tools, frameworks and test code, including scripts such as Bash
• Ability to develop software in Python or C/C++ on Linux, and to read C code to understand its behavior
• Solid Linux system fundamentals: boot process, kernel modules, device/process/memory management, networking
• Debugging skills to analyze and validate the behavior of drivers, firmware and system tools on real hardware, and to narrow down root causes based on logs and reproduction steps
• Ability to clearly explain evaluation methods and results, and to read and write technical documents and handle day-to-day work discussions in English

Preferred Qualifications

• Experience in functional and performance evaluation of PCIe devices such as NICs, DPUs, SmartNICs or AI accelerators, or an understanding of server PCIe card management functions (sensor monitoring, firmware updates, etc.)
• Experience developing or validating device drivers or management firmware
• Experience developing or validating embedded Linux/BSPs, including image builds with Yocto/OpenEmbedded, Buildroot, etc., and boot, package and update validation
• Experience with emulator-based device modeling (e.g., QEMU) or building hardware-free validation environments
• Experience building and operating test frameworks such as pytest and CI/CD with GitHub Actions or similar
• Experience building and operating a real-hardware test lab integrated with CI (device allocation, remote power control, automatic recovery, UART/JTAG/BMC/IPMI, etc.)
• Experience designing fault-injection, stress or long-duration load tests, or tracking down and reducing intermittent failures that show up across repeated runs
• Experience in performance analysis and in analyzing and visualizing test result data
• Active use of AI coding agents in development work

About us

MangoBoost, Inc.

Agentic AI is fundamentally reshaping the paradigm of AI infrastructure. As environments emerge where countless AI agents operate simultaneously, conventional data center architectures are reaching their limits in both performance and efficiency. In response to this shift, MangoBoost is pioneering a new approach with its full-stack AI infrastructure solution designed to innovate every layer of the AI data center.

From LLMBoost, our AI inference optimization software, to GPU and storage systems powered by our proprietary DPU architecture, and orchestration software that seamlessly integrates and manages the entire stack, MangoBoost designs and develops every layer of AI infrastructure in-house. Rather than building isolated products, we focus on eliminating system-wide bottlenecks and fundamentally addressing inefficiencies across the entire infrastructure stack.

MangoBoost’s technology minimizes data movement and coordination overhead while maximizing overall system resource utilization, enabling exceptional performance and optimized total cost of ownership (TCO) for customers. Today, our solutions are already being deployed within real-world AI infrastructure environments through global partners, validating the strength of our technology. Through these efforts, MangoBoost is establishing a new global standard for AI data centers.

Notice Regarding Return of Hiring Documents

• This notice is in accordance with Article 11, Section 6 of the Fair Hiring Procedure Act, which allows job applicants, excluding the final successful candidate, to request the return of any recruitment documents they have submitted.
• Job applicants who were not selected as the final candidates in the recruitment process are entitled to request the return of the recruitment documents they submitted within 180 days of the date the employment decision is confirmed. However, this does not apply to documents submitted via the website or email, or documents voluntarily submitted by the applicant without the company’s request. If the recruitment documents are lost due to natural disasters or other reasons for which the company is not responsible, it will be considered as having been returned.
• Applicants wishing to request the return of their recruitment documents should fill out the Request for Return of Hiring Documents [Form 3 of the Enforcement Rules of the Fair Hiring Procedure Act] and submit it by email ([email protected]). The documents will be sent by registered mail to the designated address within 14 days from the confirmation of submission. Please note that the cost of registered mail will be borne by the recipient.
• The company will keep the recruitment documents for 180 days following the confirmation of the employment decision. If no request for the return of documents is made within that period, the company will dispose of all recruitment documents without delay in accordance with the Personal Information Protection Act.