New opportunity

AI Software Solutions Engineer

Intel · Bangalore, India

About this role

The Role and Impact: As an AI Software Solutions Engineer, you will focus on optimizing AI solutions using Intel products, working closely with open-source frameworks like PyTorch, TensorFlow, and JAX. Your day-to-day responsibilities will involve designing and developing features for Intel's AI frameworks software stack, enabling and optimizing deep learning models, and enhancing AI frameworks for Intel's cutting-edge GPUs and accelerators. This role is pivotal to driving advancements in AI technology, ensuring Intel products deliver exceptional performance for AI workloads. Business Group: The role resides within Intel's Data Center Group (DCG), a business unit dedicated to advancing data center and AI technologies that power major innovations across industries. Key Responsibilities: Propose optimizations in SW and kernel implementation to improve workload performance. Model AI HW and SW (GPU/accelerator) for performance analysis. Analyze AI workloads to identify critical kernels and patterns. Estimate ideal performance of kernels and workloads for a given HW spec. Enhance and maintain the performance modeling toolkit for projections. Correlate projected performance with device/simulator and improve models. Study new and different DL networks for future optimization requirements. Design and develop hardware-agnostic and hardware-aware software features for AI frameworks, focusing on ML kernel development. Enhance and extend deep learning training and inference capabilities within Intel's AI software stack. Debug and resolve functional and convergence issues in multi-layered software systems. Qualifications: Minimum Qualifications: Bachelors with 8+, Masters with 6+ years, or PhD with 3+ years, of relevant industry experience in Computer Science, Computer Engineering, or Electrical Engineering. Programming in Python/C++, strong in Data Structures/Algorithms, worked on large SW projects. Knowledge of computer architecture and compiler fundamentals. Expertise in performance optimization techniques for AI workloads on modern hardware architectures. Preferred Qualifications: Familiarity with AI algorithms (e.g. LLMs) and DL and serving frameworks (e.g. PyTorch, JAX, vLLM, SGLang, HF). Advanced understanding of machine learning concepts and deep learning practices. Knowledge of Intel hardware architecture and features, including CPUs, GPUs, or AI accelerators. Experience with developing and integrating CUTLASS or Triton-based kernels in large language models (LLMs). Knowledge of compiler algorithms for heterogeneous systems and fuser optimizations. Familiarity with generative AI models and LLMs.

For full details about this role, click “Apply on company site” to view the employer’s original posting.

Skills for this role

PythonC++PyTorchTensorFlowJAXDeep LearningMachine LearningLLMsvLLMSGLangHugging FaceCUTLASSTritonComputer ArchitectureCompiler FundamentalsPerformance OptimizationData StructuresAlgorithms

Checking your profile…

Preparation suggestions, not the employer’s actual interview questions. Always check the original posting for current requirements.

Understand the role

What matters most

AI Kernel Optimization and Performance Tuning
The core of this role involves identifying critical kernels within AI workloads and implementing software optimizations to maximize performance on Intel GPUs and accelerators. You must demonstrate expertise in analyzing workload patterns and translating them into efficient hardware-aware code.
AI Framework Integration and Feature Development
You will be responsible for enhancing Intel's AI software stack by developing features for frameworks like PyTorch, JAX, and TensorFlow. This requires deep proficiency in C++ and Python to ensure seamless integration and hardware-agnostic functionality.
Performance Modeling and Projection
A key duty is maintaining and improving the performance modeling toolkit to project how AI workloads will behave on future hardware. You must be able to correlate simulation data with actual device performance to refine predictive accuracy.

Likely challenges to think through

Bridging Hardware-Software Abstraction Layers
The role requires balancing hardware-agnostic software design with the need for hardware-specific kernel optimization. Prepare by articulating how you manage the trade-offs between code portability and the low-level performance tuning required for specialized accelerators like those using Triton or CUTLASS.
Debugging Complex Multi-Layered Systems
Debugging functional and convergence issues in a stack that spans frameworks, compilers, and hardware is inherently difficult. Prepare by discussing your systematic approach to isolating performance bottlenecks or functional regressions across these disparate layers.
Adapting to Rapidly Evolving AI Architectures
The requirement to study new DL networks and LLMs for future optimization means your technical knowledge must remain current. Prepare by demonstrating your process for quickly evaluating new model architectures and identifying the specific computational patterns that will challenge current hardware limits.

Practice your answers

Technical practice

Kernel Optimization Techniques

What to coverExplain your approach to identifying and optimizing performance bottlenecks in deep learning kernels. Discuss how you use profiling tools to analyze memory access patterns, compute utilization, and potential instruction-level parallelism on modern accelerators.

Triton and CUTLASS Integration

What to coverDescribe your experience with writing or integrating custom kernels using Triton or CUTLASS. Explain the trade-offs between using high-level framework operators versus custom-written kernels for specific LLM workloads.

Compiler Fundamentals and Fuser Optimizations

What to coverHow do compiler fuser optimizations impact the performance of deep learning models? Discuss your understanding of operator fusion and how it reduces memory bandwidth overhead in heterogeneous computing environments.

Performance Modeling for AI Hardware

What to coverDescribe how you build performance models to estimate the throughput of AI workloads on specific hardware architectures. Explain how you correlate these projections with actual device measurements to refine your models.

Debugging Multi-layered Software Systems

What to coverWhen encountering functional or convergence issues in a complex AI software stack, what is your systematic debugging process? Mention how you isolate issues between the framework layer, the compiler, and the underlying hardware drivers.

Hardware-Aware Software Design

What to coverWhat are the key considerations when designing software features that need to be both hardware-agnostic and hardware-aware? Provide an example of how you would adapt a kernel to leverage specific features of an Intel GPU or accelerator.

LLM Serving Frameworks

What to coverCompare the architectural differences between serving frameworks like vLLM and SGLang. Discuss how these frameworks optimize memory management and request scheduling for large language models.

Computer Architecture for AI

What to coverHow does the memory hierarchy of modern data center GPUs influence the design of deep learning training algorithms? Suggest how you would optimize a data-heavy workload to minimize data movement.

Behavioural practice

Cross-Geo Collaboration

What to coverDescribe a situation where you had to collaborate with a team in a different geographic location to solve a complex technical challenge. How did you ensure alignment, manage time zone differences, and maintain clear communication?

Handling Technical Ambiguity

What to coverTell me about a time you were tasked with optimizing a workload where the performance requirements were not clearly defined. How did you gather the necessary information and establish a baseline for success?

Driving Technical Innovation

What to coverProvide an example of a time you proposed a new optimization or feature that significantly improved system performance. What was the impact, and how did you convince stakeholders of the value of your proposal?

Managing Large-Scale Software Projects

What to coverDescribe your experience working on large-scale software projects. How do you balance the need for rapid feature development with the requirement for long-term code maintainability and performance stability?

Sharpen your knowledge

Topics to prepare

Computer Architecture and Performance Modeling
The role requires modeling AI hardware and software for performance analysis and estimating kernel performance. Exercise: Review the memory hierarchy, cache coherence, and throughput characteristics of modern GPUs or AI accelerators. Practice calculating theoretical peak performance (FLOPS/bandwidth) for a given hardware specification and compare it against actual kernel execution times to identify bottlenecks.
Compiler Fundamentals and Fuser Optimizations
The position involves developing hardware-aware software features and understanding compiler algorithms for heterogeneous systems. Exercise: Study how operator fusion (fuser optimizations) reduces memory access overhead in deep learning frameworks. Research how compilers translate high-level graph representations into optimized machine code for specific hardware backends.

Skills to revise

ML Kernel Development with Triton or CUTLASS
The role explicitly mentions developing and integrating kernels for LLMs. Exercise: Write a custom matrix multiplication or activation function kernel using Triton or CUTLASS. Focus on optimizing memory tiling, shared memory usage, and thread-level parallelism to achieve high hardware utilization.
Deep Learning Framework Optimization (PyTorch/JAX)
You will be enhancing AI frameworks for Intel hardware. Exercise: Profiling a standard LLM inference workload using PyTorch or JAX to identify compute-bound versus memory-bound kernels. Use profiling tools to inspect kernel execution traces and propose specific code changes to improve throughput.
C++ and Python for Large-Scale Software Systems
The role requires working on large software projects with a focus on performance. Exercise: Practice implementing high-performance data structures or algorithms in C++ that minimize latency. Ensure your code follows modern C++ standards and demonstrates efficient memory management, which is critical for low-level framework development.

Your preparation checklist

Review AI Kernel Optimization Techniques
Prepare to discuss your experience with performance tuning for deep learning workloads. Focus on your ability to identify bottlenecks in PyTorch or JAX, and be ready to explain how you have utilized or integrated Triton or CUTLASS kernels to improve execution speed on modern hardware.
Prepare a Technical Walkthrough of a Large Software Project
Select a past project that involved complex software architecture or compiler-related tasks. Be prepared to explain your design choices, how you handled hardware-aware optimizations, and the specific metrics you used to measure performance improvements.
Refresh Computer Architecture and Compiler Fundamentals
Review core concepts such as memory hierarchy, cache optimization, instruction sets, and heterogeneous system compilation. Be ready to apply these concepts to explain how software features can be optimized for specific GPU or accelerator architectures.
Develop a Performance Modeling Case Study
Draft a brief explanation of how you would approach modeling a new AI workload for a specific hardware target. Include your methodology for identifying critical kernels, estimating theoretical performance limits, and correlating those projections with actual device measurements.
Formulate Strategic Questions for the Interviewers
Prepare questions regarding the current challenges in Intel's AI software stack, the team's workflow for cross-geo collaboration, or how the group balances hardware-agnostic development with the need for hardware-specific performance tuning.

Availability has not yet been verified. Status can change after a check; confirm on the employer’s website. Report an issue · Our checking policy

KEEP EXPLORING

Similar AI jobs

Related skills and specializations in India. Matched to this role, not your profile.

Explore more jobs

Finding similar opportunities…