ML/AI Engineer - CAD Infrastructure

The main responsibilities of the Machine Learning/AI Engineer at Tenstorrent revolve around developing ML-powered systems for automated post-silicon validation, maintaining the infrastructure for large-scale silicon characterization and testing, and creating predictive models to analyze silicon behavior. You will also implement AI-driven solutions for anomaly detection and build ML pipelines for processing extensive datasets related to post-silicon validation.

To qualify for the Machine Learning/AI Engineer position at Tenstorrent, candidates should hold a Master’s degree or Ph.D. in Computer Science, Electrical Engineering, or a related field, along with a minimum of 5 years of experience in post-silicon validation or CAD infrastructure development. Strong backgrounds in machine learning frameworks, proficiency in Python, and a good understanding of computer architectures and validation methodologies are also essential.

At Tenstorrent, success for a Machine Learning/AI Engineer is defined by the engineer's ability to innovate and improve the post-silicon validation pipeline through the application of AI and machine learning techniques. This includes successfully developing ML models that enhance validation efficiency, accurately diagnosing silicon issues, and streamlining testing processes—all while effectively collaborating with cross-functional teams.

The Machine Learning/AI Engineer at Tenstorrent will work with a rich toolkit, including machine learning frameworks like PyTorch and TensorFlow, data analysis libraries such as NumPy, Pandas, and Scikit-learn. Other tools may include debugging equipment like JTAG and ILA as well as programming languages such as Python, C++, and shell scripting for automation.

Tenstorrent fosters a work culture centered on collaboration, innovation, and a shared passion for solving complex problems. As part of the CAD Infrastructure team, the Machine Learning/AI Engineer can expect to work alongside talented professionals who are driven by curiosity and a commitment to excellence, all within a flexible hybrid work environment that values each employee's contributions.

To develop an ML-powered system for silicon validation, I would first identify key validation metrics and gather historical data for training. Next, I would select appropriate machine learning algorithms to model silicon behavior, ensuring that I preprocess the data effectively. I would then validate the model using a separate dataset, iteratively refining it to improve accuracy, and finally integrate it into the existing validation framework for automated deployment.

I have comprehensive experience with post-silicon validation methodologies, notably through hands-on roles that involved silicon bring-up and characterization. This includes working on various validation projects where I applied standard methodologies while incorporating ML for anomaly detection, thereby enhancing the overall validation process for higher efficiency.

In addressing anomaly detection during silicon bring-up, I would leverage machine learning models trained on historical behavior data to identify patterns. I would implement real-time monitoring systems that track various performance indicators. Any deviations from the norm would trigger alerts, allowing for quick root cause analysis, thus mitigating potential issues proactively.

I am most familiar with ML frameworks like TensorFlow and PyTorch. I have applied TensorFlow in building neural networks for predictive modeling and used PyTorch for rapid prototyping and implementation of deep learning techniques. My experience encompasses everything from data preprocessing to deploying models in production settings.

One challenging problem I encountered in CAD infrastructure development involved inefficiencies in data processing during silicon validation. I addressed this by designing streamlined ML pipelines that automated many of the time-consuming data analysis tasks, significantly reducing processing time and improving the accuracy of our validation outputs.

To keep my ML models accurate over time, I implement continuous training and validation processes. This includes setting up pipelines that automatically update the models as new data is obtained during silicon testing. I also regularly assess performance metrics, ensuring models adapt to any changes in silicon behavior or validation criteria.

I am proficient in Python, C++, and shell scripting. Python is my primary language for developing machine learning models and analysis scripts, while I utilize C++ for performance-critical parts of our validation tools. Shell scripting helps automate various tasks and streamline workflows, contributing to overall productivity.

I approach collaboration with cross-functional teams by maintaining open and regular communication, ensuring team members are aligned on project goals. I also actively engage during brainstorming sessions, contributing my expertise while being receptive to ideas and input from other specialists, thereby fostering a highly collaborative environment.

For handling large datasets, I prefer using data analysis tools like Pandas and NumPy in Python due to their powerful features for data manipulation and performance. Additionally, SQL is essential for database queries and optimizing data retrieval, ensuring efficient analysis and processing, especially in validation projects.

I am motivated by the transformative potential of AI and machine learning in various fields, particularly in semiconductor technology. The ability to solve real-world challenges through innovative solutions and to push the boundaries of what is possible in silicon validation drives my passion, making each project an exciting opportunity for creativity and growth.