Debabrata Acharjee

Software Engineer

Bengaluru, Karnataka, India8 yrs 8 mos experience

Most Likely To SwitchAI ML Practitioner

Key Highlights

Expert in GPU and graphics driver optimization.
Proven track record in AI infrastructure development.
Strong background in Linux and Windows driver stacks.

Stackforce AI infers this person is a GPU and AI infrastructure specialist with deep expertise in driver development.

Contact

Skills

Core Skills

Graphics Processing UnitC++Linux Kernel

Other Skills

Linux GPU DriverAI InfrastructureGPU virtualizationPerformance TuningPythonAnsibleCI/CDGPU display pipelinesMultimedia accelerationPower managementPerformance benchmarkingCI-driven automationDebuggingAI MLDirectX

About

I focus on GPU, CPU and Graphics Driver and what is the most that can be obtained out of them. Specialization: - Knowledge on Windows OS /Linux(debian, Ubuntu) internals - KMS / DRM / Wayland / Western / X - Win32API - WIN 32 API - OpenGL - Language - C / C++ / Python - Scripting: VBscript , Python, Bash Scripting, batch file scripting. - Competitve Performance Benchmarking - Platform/Hardware Validation - Software Validation - Performance Tuning - Windows Device Driver QA - Linux Graphics driver QA - WHQL Certification for windows/Linux Graphics Driver.

Experience

8 yrs 8 mos

Total Experience

2 yrs 10 mos

Average Tenure

4 yrs 11 mos

Current Experience

Intel corporation

Graphics Software Development Engineer

Jun 2021 – Present · 4 yrs 11 mos · Bengaluru · Hybrid

Linux GPU Driver & AI Infrastructure
Architected and optimized Intel i915 GPU driver for SR-IOV and PCIe passthrough, enabling production-grade GPU virtualization across data center and edge platforms.
Delivered end-to-end GPU virtualization (KVM/QEMU/libvirt) from firmware bring-up to hypervisor integration, enabling multi-tenant AI inference and training environments.
Brought up and performance-tuned LLM and deep learning workloads (PyTorch, TensorFlow, ONNX Runtime) on virtualized and bare-metal Linux systems.
Reduced AI inference latency by optimizing GPU memory residency, context switching, interrupt paths, and submission queues within the DRM stack.
Enhanced scheduler, GT reset, and watchdog flows to improve stability under large tensor workloads (>20GB), supporting transformer training and mixed AI + rendering use cases.
Designed multi-GPU validation platforms to evaluate distributed training scalability, resource contention, and heterogeneous GPU behavior.
Built Python/Ansible automation for SR-IOV and AI infrastructure validation, enabling CI/CD pipelines and significantly reducing regression turnaround time.
Partnered with AI framework, kernel, and hypervisor teams to align driver optimizations with ML runtime execution paths and distributed compute architectures.
AI Systems & Acceleration
Engineered GPU-accelerated platforms supporting concurrent 4K rendering and on-device LLM inference with deterministic QoS.
Implemented zero-copy memory pipelines and telemetry systems tracking SM utilization, bandwidth, latency, and scheduler fairness for AI workloads.
Infrastructure & QAT
Built fleet validation and node certification tooling for large IaaS GPU deployments, ensuring firmware, driver, and AI runtime readiness.
Enabled Intel QAT with SR-IOV, crypto, and compression acceleration; automated regression testing and resolved kernel-space stability issues for production releases.

Linux GPU DriverAI InfrastructureGPU virtualizationPerformance TuningPythonAnsible+2

Amd

Software Engineer

Feb 2018 – Jun 2021 · 3 yrs 4 mos · Bengaluru

Parent: Tech Mahindra Pvt Ltd.
GPU Display & Graphics Driver Expertise
Deep expertise in GPU display pipelines across DP, HDMI, DVI, and VGA, including MST (daisy chain), SLS/Eyefinity, True 4K, HDCP, FreeSync, VSync, and multi-GPU rendering.
Debugged and optimized display bring-up flows covering link training, EDID handling, hot-plug detection, bandwidth allocation, and multi-monitor topology stability.
Delivered stable multi-4K and high-refresh deployments across discrete and integrated GPUs, resolving timing, flicker, and synchronization issues across heterogeneous platforms.
Strong experience in multimedia acceleration: H.264/H.265/VP9 playback & transcode, HDR pipelines, PlayReady compliance, and seamless playback across SLS and multi-display setups.
Enabled and validated Vulkan, OpenGL, DirectX, and OpenCL workloads, ensuring graphics and compute stability across Windows and Linux driver stacks.
Worked extensively on power management flows (S3, S4, reboot, stress), debugging GPU resets, context recovery, and display reinitialization issues.
Hands-on with PCIe, eMMC, NVMe, SPDK, and high-speed networking (1/10GbE, DPDK) in performance-sensitive GPU systems.
Conducted performance benchmarking using Unigine, Cinebench, and PassMark to analyze rendering throughput, memory bandwidth, and display stability under load.
Experienced in low-level IO interfaces (GPIO, I2C, UART, I2S, PCIe) for board bring-up and hardware-software integration.
Utilized graphics and compute profilers, vendor tools, and kernel tracing for root-cause analysis of rendering, display, and memory bottlenecks.
Implemented CI-driven automation (Jenkins, Python, Shell) across APUs and dGPUs, reducing regression cycles and improving driver release quality.
Platforms: Linux (Ubuntu), Windows (Win7/10), Hypervisors (KVM, ESXi, Citrix), supporting thin client and virtualized graphics deployments.

GPU display pipelinesMultimedia accelerationPower managementPerformance benchmarkingCI-driven automationLinux Kernel+1

North eastern regional institute of science and technology (nerist)

Assistant Professor (Adhoc)

Jul 2016 – Dec 2016 · 5 mos · Arunachal Pradesh, India

My responsibilities includes delivering lectures to allotted classes on engineering subjects, taking allotted Lab classes , evaluating students etc on behalf of Dept. of ECE besides doing other departmental works.