Compute In-Memory System Engineer Intern
Change the world. Love your job.
At TI, Systems engineers focus deeply on understanding the technical needs, and future trends of an industry or end equipment, then create new products and innovative forward-looking product roadmaps to solve them. Systems Engineers are an integral part each phase of new product development at TI. In the early stages of product development, systems engineers interface with key stakeholders (customer decision-makers, application engineers, marketing, management, sales, IC design engineers, technology development) to negotiate specifications, perform trade-offs, understand the competitive landscape, and ultimately develop detailed technical definitions for new products. They then collaborate with the full IC development team (design, applications, test, product engineers) to deliver products to the market which are compelling, competitive, cost-conscious, manufacturable, and importantly, successful in growing TI’s business.
We are seeking a highly motivated PhD student to join our Embedded AI team in Kilby Labs this summer to work on cutting-edge digital compute in-memory (CIM) research and development for resource-constrained microcontroller applications. As a key member of our team, you will focus on advancing energy-efficient computing architectures that enable intelligent edge processing on cost and power-constrained platforms, including architectures that tile and scale digital CIM to deliver much higher compute throughput. Your work will involve exploring innovative approaches to co-optimize algorithms, architectures, and hardware implementations for CIM systems.
In this systems engineering intern role, you’ll have the chance to:
Research and develop novel digital compute in-memory (CIM) architectures optimized for edge AI workloads on resource-constrained microcontrollers
Explore tiling of digital CIM macros, ganging multiple CIM tiles together to scale compute throughput well beyond a single macro while staying within strict area, power and cost budgets
Study inter-tile architecture trade-offs: dataflow, workload partitioning and mapping across tiles, tile-to-tile communication, scheduling, and utilization
Quantify throughput, energy efficiency (TOPS/W) and area scaling of tiled CIM versus conventional NPU/MAC-array approaches, including the impact of mixed-precision and sub-8-bit quantization
Explore algorithm-hardware co-design techniques to maximize energy efficiency and computational throughput
Collaborate with internal technology and design teams to simulate and validate CIM-based solutions
Develop and optimize machine learning models and inference flows tailored for in-memory computing paradigms
Participate in the design and implementation of software toolchains and simulation frameworks for CIM and tiled-CIM evaluation
Investigate memory hierarchy optimizations and data movement reduction strategies for edge processing
Document findings and present trade-offs and recommendations to architecture, design and business stakeholders
Minimum Requirements:
Currently pursuing a PhD in Electrical Engineering, Computer Engineering, Electrical and Computer Engineering or related field
Cumulative 3.0/4.0 GPA or higher
Preferred Qualifications:
Solid background in computer architecture, digital design, memory systems, and/or embedded systems
Research experience in compute in-memory (CIM), processing-in-memory (PIM) or near-memory computing, ideally demonstrated by publications at leading architecture/circuits/systems venues (e.g., ISCA, MICRO, HPCA, DAC, ISSCC)
Familiarity with scalable / multi-tile accelerator architectures, including dataflow, mapping and scheduling of neural networks across multiple compute tiles
Proficiency in hardware description languages (Verilog/VHDL) and/or architectural simulation tools
Strong programming skills in Python and C/C++, including performance analysis and optimization
Experience with machine learning model optimization, quantization (including sub-8-bit), and deployment on resource-constrained devices
Knowledge of SRAM/DRAM circuit design and memory architecture principles
Familiarity with deep learning frameworks (e.g., PyTorch, TensorFlow) and embedded ML libraries (e.g., TensorFlow Lite, CMSIS-NN)
Understanding of energy-efficient computing techniques and power/area modeling
Experience with algorithm-hardware co-design methodologies
Ability to establish strong relationships with key stakeholders critical to success, both internally and externally
Strong verbal and written communication skills to audiences of varied background
Ability to simplify complex problems and navigate uncertainty
Ability to quickly ramp on new systems and processes
Demonstrated strong interpersonal, analytical and problem-solving skills
Ability to work in teams and collaborate effectively with people in different functions
Ability to take the initiative and drive for results
Strong time management skills that enable on-time project delivery
