Tech Lead Software Engineer, ML (Unified Model)

As a Tech Lead Software Engineer at Latitude AI, you will be responsible for developing cutting-edge spatio-temporal machine learning models, facilitating multi-sensor fusion, and preparing solutions for real-world implementation. You will also collaborate with perception experts to prototype and test algorithms, while maintaining high software standards throughout the project lifecycle.

Candidates for the Tech Lead Software Engineer position at Latitude AI should have a Bachelor's degree in a related field, coupled with significant experience in machine learning and deep learning solutions. A Ph.D. or equivalent experience is preferred, and expertise in working with technologies such as PyTorch and Tensorflow is essential.

Latitude AI emphasizes professional growth for its Tech Lead Software Engineers through various initiatives, including reimbursement for professional development courses, which allows you to further enhance your skills, stay current with industry trends, and ultimately contribute more significantly to our goal of reimagining transportation.

The culture at Latitude AI is collaborative, innovative, and fast-paced. As a Tech Lead Software Engineer, you will be part of a diverse team that fosters creativity and open communication, encouraging everyone to contribute ideas while working towards groundbreaking solutions in autonomous driving technology.

As a Tech Lead Software Engineer at Latitude AI, you will work with advanced tools and technologies, including machine learning frameworks like PyTorch and Tensorflow, as well as various sensor technologies like Camera, Radar, and LiDAR. You will also engage with development environments in Python and C++.

When developing machine learning models for real-time applications, I start by understanding the specific requirements and constraints of the application environment. It's essential to collect relevant datasets, ensure data quality, and then progressively build and test models, emphasizing optimization for speed and efficiency. Collaborating with cross-functional teams for feedback and challenges is also crucial.

In my previous roles, I've extensively used PyTorch for its flexibility in model building and debugging, particularly in developing custom neural network architectures. Tensorflow has been beneficial for its production-ready deployment capabilities, especially when shipping models to various platforms. Familiarity with both frameworks allows me to choose the best one based on project needs.

At my last job, we developed a model for object detection that performed well in the lab. To transition it into production, I led a team to rigorously test the model in simulated environments, gathered feedback, and iterated on our design. We then moved to real-world tests, ensuring all the necessary safety parameters were met prior to deployment. The model is now successfully embedded in consumer products.

To optimize machine learning algorithms, I analyze their performance metrics and implement strategies such as hyperparameter tuning, reducing model complexity without sacrificing accuracy, and leveraging techniques like transfer learning. Regularly reviewing and staying updated on the latest research also informs new optimization techniques.

I believe in open communication and setting clear expectations from the beginning. I regularly engage with perception experts and roboticists through brainstorming sessions, project meetings, and feedback loops to navigate technical challenges and integrate our findings effectively into solutions. Collaboration is key to our success.

I have worked on several projects involving multi-sensor fusion, where data from different sources such as cameras, LiDAR, and radar are combined to improve perception accuracy. I focus on developing algorithms that efficiently process and unify these data streams, enhancing robustness and providing richer information for downstream decision-making.

To stay updated on advancements in machine learning and computer vision, I regularly read research papers from top conferences, participate in webinars, and attend industry workshops. Joining online communities also allows me to engage with peers, share insights, and discuss emerging technologies and methodologies.

Developing perception systems often presents challenges such as data quality issues and model generalization. I address these by implementing robust validation techniques, collecting diverse datasets, and fostering collaboration to brainstorm solutions. Continuous testing and feedback help in iterating on our designs effectively.

I have successfully delivered multiple computer vision software products by following best practices in Agile development. My experience involves gathering user requirements, iterating on product designs based on user feedback, rigorous testing, and ensuring that deployment pipelines are smooth and efficient for seamless integration into existing systems.

Success in the Tech Lead Software Engineer role at Latitude AI involves not only developing high-performing models but also fostering a positive team culture, driving innovations that push the boundaries of technology, and ensuring that our solutions positively impact real-world applications, making driving safer and more enjoyable.