ML Analyst - Adversarial AI
Job ID: 358
As a Two Six Technologies ML Analyst, you will join our adversarial AI applied research project team. You will analyze and red-team machine learning systems and defenses and evaluate their robustness to adversarial attack. You will apply techniques from machine learning, numerical optimization, simulation rendering, model hooking, and statistics to bypass insecure defenses with adaptive attacks and measure true model robustness. You will be a primary user of our Armory open source framework github.com/twosixlabs/armory designed for model robustness evaluation, and you will provide QA feedback to our team, provide feature requests, develop new analytical features, and help develop tutorials and documentation for effective use. You may also develop algorithmic tools to improve the effectiveness of Armory or lightweight front-ends, such as streamlit in Jupyter, to display and communicate evaluation results.
Duties:
- Evaluate the robustness of ML models and defenses to adversarial attacks, particularly inference-time evasion and data poisoning, on various tasks including object detection, object tracking, speech-to-text, and speaker identification.
- Analyze models and defensive techniques to find novel adaptive attacks and combinations of techniques that are tuned to the particular ML system being evaluated.
- Write and communicate reports and/or notebooks of model evaluations, including your methodology and results.
- Provide quality assurance (QA) feedback including issues and bug reports, provide feature requests, provide test requirements, and develop documentation for the Armory codebase.
- Assist with developing tutorials for Armory use, algorithmic tools for improving evaluations, or lightweight front-ends for interactively displaying results.
- Work collaboratively with peers, including software developers and research scientists, in an agile environment on a fully-remote team.
Requirements:
- 1+ year Python development experience
- BS degree in CS, EE, math, or similar field, or equivalent experience
- Strong applied math fundamentals, such as numerical optimization
- Familiar with common machine learning pipelines, model architectures, and metrics
- Comfortable reading and analyzing academic publications in machine learning
- Comfortable writing, discussing, and communicating results with the team
Nice-to-haves:
- PyTorch, TensorFlow, or JAX - experience building, training, and/or analyzing models
- Advanced degree in CS, EE, math, or similar field, or equivalent experience
- Open source software - experience with git, agile development, CI testing, code reviews, and understanding other people’s code
- Experience with Jupyter, streamlit, or similar interactive machine learning tools
Clearance Requirements:
- Eligibility for Security clearance
#LI-JS1
#LI-Remote
Two Six Technologies is an Equal Opportunity Employer. All qualified applicants will receive consideration for employment at Two Six Technologies without regard to race, color, religion, national origin, sex, age, physical and mental disability, sexual orientation, gender identity or expression, genetic information, veteran, marital, pregnancy or citizenship status; or any other status prohibited by applicable national, federal, state or local law.
Two Six Technologies Covid-19 Vaccination Policy requires employees to be fully vaccinated. Exceptions to this policy are only granted to those with a company-approved medical or religious accommodation. Prospective or new employees will be required to adhere to this policy and submit proof of vaccination or have an approved exemption prior to the start of their employment.
If you are interested in applying for employment with Two Six Technologies and require an accommodation, please contact Human Resources at Two Six Technologies by sending an email to [email protected]. Information provided will be kept confidential and used only to the extent required to provide needed reasonable accommodations.