Tascent’s engineering team is creating a new generation of low-power embedded biometric recognition hardware products. Currently, Tascent has openings for individuals adept in computer vision to develop a variety of algorithms designed to operate on low-power processing hardware. In addition, the role will also involve development and evaluation of uncompressed, higher-performing algorithms designed for embedded x86, ARM, and mobile platforms and non-embedded cloud platforms. The AI Engineer will report to our Lead AI Engineer, and will work closely with our product development team.
Key responsibilites
- Work with the Tascent Engineering team to develop biometric detection, recognition, presentation attack detection, and other related algorithms on low-power devices.
- Write requirements for validation and training datasets to minimize demographic, race-specific, and other biases.
- Work with data engineers to characterize model outliers, generate performance status reports, and develop solutions to challenging AI problems.
- Comply with the ethical and legal handling of PII, including consent management within relevant datasets.
Qualifications
- A Bachelor’s degree (min) or ideally a MSc or PhD in computer science, engineering, mathematics, or other technical degrees with an emphasis in ML/AI (or equivalent experience).
- 2+ years developing vision-based AI models in a product development environment.
- Familiarity with sources of bias in data collection and AI model development.
- Demonstrated understanding of AI model compression techniques, and ideally hands-on experience preparing models for more than one AI accelerator hardware offering.
- Applied understanding of GANs.
- Knowledge of python + keras/tensorflow (alternatively pytorch or MATLAB/R).
- Experience with models targeted for edge devices / CPU.
- Experience with any MLOps platform (i.e. weights and biases).
- General knowledge of best practices pertaining to product-level software development.
Apply for this job
Email a resume and letter of interest to [email protected]