Lead Engineer (HPC, GPU, CUDA)

AIRA Matrix; Thane; On-site; INR 1,800,000 - 3,200,000; Full time

Job summary

AIRA Matrix in India is seeking an AI/ML Architect to define architectural changes and accelerate deep learning models across GPU/CPU stacks, translating business needs into production-ready goals.

You will benchmark CV algorithms, optimize performance, and drive end-to-end workflows from data curation to deployment, collaborating with cross-functional teams to deliver scalable AI solutions.

Qualifications

• Bachelor-level degree in CS/EE or related field.
• Strong background in deployment of complex deep learning architectures.
• 1+ years in ML/DNN with focus on DL fundamentals and model optimization.
• Proficient in C++ with knowledge of data structures and algorithms.
• Experience with DNN frameworks (Torch, Caffe, TensorFlow).
• Proficient in Python and bash scripting.
• Experience with Windows, Ubuntu and Centos.
• Excellent communication and collaboration skills.

Responsibilities

• Identify architectural changes or new approaches to accelerate deep learning models.
• Translate business needs into product goals for AI-ML workloads and deployment.
• Benchmark and optimize CV algorithms on heterogeneous hardware (GPU/CPU).
• Collaborate to drive end-to-end workflow from data curation to deployment.

Skills

Deep learning architectures

DNN deployment

C++ programming

Python scripting

Performance optimization

DL fundamentals

Communication skills

Education

Bachelor's or higher in CS/EE

Tools

CMake

Make

Ninja

Clang-Tools

TensorRT

CuDNN

PyTorch

CUDA

OpenCL

OpenMP

MPI

Docker

Singularity

Kubernetes

SLURM

LSF

Job description

Responsibilities

• We seek an expert to identify architectural changes and/or completely new approaches for accelerating our deep learning models.
• As an architect you are responsible for converting business needs associated with AI-ML algorithms into a set of product goals covering workload scenarios, end user expectations, compute infrastructure and time of execution; this should lead to a plan for making the algorithms production ready.
• Benchmark and optimize the Computer Vision Algorithms for performance and quality KPIs on the heterogeneous hardware stacks (GPU + CPU, etc.).
• Collaborate with various teams to drive an end to end workflow from data curation and training to performance optimization and deployment.

Skills Required

• Bachelors or Higher in Computer Science, Electrical Engineering, or related field.
• A strong background in deployment of complex deep learning architectures.
• 1+ years of relevant experience in at least a few of the following relevant areas: Machine learning (with focus on Deep Neural Networks), including understanding of DL fundamentals; Experience adapting and inferencing DNNs for various tasks; Experience developing code for one or more of the DNN training frameworks (such as Torch, Caffe or TensorFlow): Numerical analysis, Performance analysis, Model compression and Optimization & Computer architecture.
• Strong data structures and algorithms knowhow with excellent modern C++ programming skills.
• Good grasp over software engineering and tools like CMake, Make (or Ninja), Clang-Tools, etc.
• Hands-on expertise with TensorRT, CuDNN, PyTorch.
• Hands-on expertise with GPU computing (CUDA, OpenCL or OpenACC) and HPC (MPI, OpenMP).
• Proficient in Python programming and bash scripting.
• Proficient in Windows, Ubuntu and Centos operating systems.
• Excellent communication and collaboration skills.
• Self-motivated and able to find creative practical solutions to problems.

Good to Have

• Hands-on experience with PTX-ISA for CUDA or vector intrinsics like AVX, SSE, etc.
• In-depth understanding of container technologies like Docker, Singularity, Shifter, Charliecloud.
• Hands-on experience with HPC cluster job schedulers such as Kubernetes, SLURM, LSF.
• Familiarity with cloud computing architectures.
• Hands-on experience with Software Defined Networking and HPC cluster networking.
• Working knowledge of cluster configuration management tools such as Ansible, Puppet, Salt.
• Understanding of fast, distributed storage systems and Linux file systems for HPC workloads.