Adesh Gupta

WORK HISTORY

EDUCATION

ABOUT ADESH GUPTA

Experienced Embedded AI Engineer with deep expertise in AI model optimization, embedded systems, and low-level debugging. Passionate about deploying high-performance AI models on edge devices while solving complex challenges across Linux kernel, device drivers, and SoC platforms. Embedded AI Deployment – Fine-tuned transformer models (BERT, T5, DistilBERT) for NLP tasks and deployed them on ARM Cortex-A boards using TensorFlow Lite and ONNX Runtime. Applied quantization and pruning to reduce model size by 40% with minimal accuracy loss. SoC Debugging & Validation – Hands-on experience debugging boot failures, kernel panics, and crash dumps using JTAG, T32 Debugger, and kernel log tracing. Improved system uptime and reliability across embedded platforms. Linux Kernel & Drivers – Developed and optimized Linux kernel modules and device drivers for custom peripherals, enhancing hardware-software interaction and system throughput. AI Optimization & Inference – Converted models to ONNX and TFLite formats, achieving sub-200ms inference latency on constrained hardware. Explored compiler-level optimizations using TFLite delegates and NNAPI. Hardware-Software Co-Design – Collaborated across compiler, firmware, and system teams to trace low-level execution paths and optimize boot-time performance. Skilled in C/C++, Python, and embedded firmware development.Key Skills: AI Model Optimization – Quantization, Pruning, ONNX Runtime, TFLite Embedded Systems – ARM Cortex-A/M, SoC Architecture, Linux BSP Kernel Debugging – Ftrace, GDB, Perf, Crash Analysis AI Frameworks – TensorFlow, Keras, Scikit-learn, Transformer Models Tools & Compilers – Git, Docker, Make/CMake, TFLite Converter, ONNX Graph OptimizerPreviously worked on deploying NLP models for aviation and hate speech detection, solving real-world bottlenecks in performance, memory, and latency. Passionate about edge AI, embedded ML, and next-gen compute architectures.Let’s connect if you\'re working on AI at the edge, embedded systems, or Linux kernel performance!