Looking for a job

Mehar vohra

@meharvohra

Message

AI/ML engineer at IHP who extended an NVDLA simulator for multi-bit fault injection in neural networks.

Germany

Message

At IHP – Leibniz Institute for High Performance Microelectronics, I built Python and PyTorch pipelines to train and evaluate DNNs, including AlexNet, ResNet-9, and VGG-11.

I extended an NVDLA-based fault-injection simulator to support simultaneous multi-bit faults, then ran statistical campaigns to study error rates and corruption severity across network architectures.

My projects include an end-to-end RAG system with ChromaDB and FastAPI, an agentic research system with tool calling, and a framework for benchmarking open-source LLMs.

Experience

Work history, roles, and key accomplishments

Research in AI-Dedicated Software Development

IHP – Leibniz Institute for High Performance Microelectronic

May 2024 - Apr 2026 (1 year 11 months)

Built and maintained Python and PyTorch training and evaluation pipelines for multiple DNN architectures, including dataset preparation, model training, validation, and result tracking. Implemented DNN quantization workflows and extended an NVDLA-based fault-injection simulator for multi-bit-flip fault injection.

Python PyTorch DNN Training Model Quantization Fault Injection Testing NVDLA C++PyBind Git Linux

Education

Degrees, certifications, and relevant coursework

Brandenburg University of Technology

Master of Science, Artificial Intelligence

Pursuing an M.Sc. in Artificial Intelligence with expected completion in September 2026. Coursework includes Data Mining, Explainable Machine Learning, Causal Data Science, AI/ML for Data Exploration, and Statistics.