Daniel Hacobian

• Conduct AI safety research on hidden misalignment in language models, studying whether deceptive behavior can be detected through internal activation patterns.
• Build activation-based diagnostic tools using PyTorch, Hugging Face model hooks, and the bench-af framework to extract and classify model representations.
• Train linear and non-linear probes on honest vs. deceptive model outputs, analyzing layer-wise patterns and cross-model generalization across GPT-2, HAL9000, and LLaMA-3-70B.

• Teach programming fundamentals through Scratch, JavaScript, and C# with Unity in classroom and camp settings.
• Guide students through debugging, problem-solving, and project development while encouraging creativity and independent thinking.

• Lead weekly virtual sections for students learning introductory programming through Python and Karel.
• Support students with debugging, problem-solving, and coding exercises while fostering a welcoming learning environment.
• Explain core CS topics — variables, loops, conditionals, functions, and decomposition — in a clear, beginner-friendly way.

• Installed, wired, and tested low-voltage automation systems for residential and commercial projects, ensuring functionality, safety, and reliable performance.
• Used multimeters, oscilloscopes, and wiring diagrams to diagnose circuit issues, repair faulty connections, and verify system operation.
• Configured control panels, sensors, relays, and power supplies while supporting hardware integration, testing, and on-site setup.

• Led the electronics and pneumatics subteam for an FRC competition robot, overseeing wiring, sensor integration, pneumatics, and electrical board organization.
• Designed, assembled, and maintained robot electrical systems including motor controllers, sensors, solenoids, compressors, and low-voltage control components.
• Troubleshot electrical and pneumatic issues during build season, testing, and competitions to improve reliability and reduce downtime.
• Collaborated with mechanical, programming, and drive teams to integrate hardware systems and prepare for regional and national competitions.

A mobile app that scans barcodes and identifies allergen risks to help users make safer food choices. Placed 3rd of 68 in the Congressional App Challenge (Rep. Judy Chu's district).

Incoming Transfer Student

Relevant coursework: Data Structures, C/C++, Discrete Structures, Linear Algebra, Honors Multivariable Calculus, Honors Differential Equations, Physics for Engineers (Mechanics, E&M, Waves/Optics), Statistics.

Founder & President — Machine Learning Club

Founded and led a student club focused on teaching core ML concepts through hands-on projects, organizing lessons and guiding members through building models.