Senior Automotive Systems Engineer
Working at the intersection of automotive SoCs, on-device AI, and silicon performance engineering to enable next-generation platforms.
I'm an Automotive Systems Engineer at Qualcomm, working across automotive SoCs, on-device AI, and system performance engineering. My focus is delivering reliable real-world performance for next-generation automotive platforms.
My work spans ARM and Nuvia CPU architectures, system benchmarking, real-world concurrent use-case enablement, and LLCC/system-cache analysis. I enjoy solving complex system problems, building tools that cut engineering cycles, and enabling advanced workloads: multi-camera pipelines, GenAI, and LLM on constrained embedded platforms.
Outside engineering, I'm drawn to deep intellectual exploration: philosophy, political ideas, and historical systems. I value creativity, self-reflection, and long-term personal growth.
Building and optimising systems where hardware meets software at the edge of real-world performance.
CPU, GPU, DDR level performance analysis on ARM and Nuvia (Oryon) architectures for next-gen automotive silicon.
Profiling and optimising MLPerf and GenAI workloads (LLMs, multi-camera) on constrained embedded platforms using QNN SDK.
Designing low-level benchmarking pipelines, VMIN characterisation, and performance dashboards for silicon validation.
Real-world concurrent use-case analysis and LLCC/system-cache tuning for complex automotive platform workloads.
A selection of projects spanning systems automation, performance engineering, and applied ML.
Designed and built a Python-based automation system to compute silicon Voltage Minimum (VMIN) for Qualcomm SoC characterisation. Reduced a previously 7-month manual process to approximately 1 month through automated test orchestration and data analysis pipelines. Integrated directly into the SoC bring-up workflow enabling faster iteration on CPU performance validation.
Profiled and optimised MLPerf inference workloads running on Qualcomm ADAS SoC platforms using the Qualcomm Neural Network (QNN) SDK. Identified micro-architecture bottlenecks and proposed optimisation strategies to improve throughput and latency for automotive AI workloads.
Implemented a space-efficient probabilistic data structure (Bloom Filter) to check user existence without false negatives in a vaccination registry system. Used MurmurHash3 (mmh3) as the primary hashing function for high performance and low collision rates.
Technical experiments leveraging agentic AI to push the boundaries of research and visualisation.
An experimental exploration into cognitive patterns and historical systems. Visualized via a bespoke 3D/2D interface designed to reveal latent connections within high-dimensional datasets.
Launch Environment →Open to conversations about systems engineering, automotive AI, or anything interesting.
arjitc12@icloud.com