FORTIS Lab

FORTIS stands for Foundations Of Reliable and Trustworthy Intelligent Systems. Drawing on the Latin word fortis — “strong” or “resilient” — the FORTIS Lab develops AI systems that remain reliable under uncertainty, safe in complex environments, and scalable in real-world deployment. Our research spans two core dimensions: advancing the foundations of reliability and safety in modern AI models, and building open, scalable systems and scientific applications that translate these foundations into practical impact. This includes work on out-of-distribution and anomaly detection, LLM and agent safety, and graph and multimodal intelligence. Through open-source tools and interdisciplinary collaboration, FORTIS aims to deliver AI that is rigorous, dependable, and broadly beneficial to science and society.

1. Openings and Candidate Fit I may recruit Ph.D. student for Fall 2027 -- it will be very different due to the funding constraints (see below). We also have openings for undergraduate and graduate interns all the time, both at USC and elsewhere (preferably in North America for time zone compatibility).

• Research Focus: Candidates should align with my core research areas. For details, please visit my research interests.
  Update (Jan 2026): For Fall 2027, due to funding reasons, my lab will rely on fellowship offers but no RA offers (which require an impossible amount of funding). Thus, this year we can only recommend candidates for fellowship but the fellowship committee will make the final decision. This requires applicants to have high GPAs (3.7+) and reasonably good English test scores (TOEFL 100+) for committee review. Of course, a good number of papers are also expected.
• Prerequisites for Ph.D.: Successful candidates typically have:
  • A publication record comparable to our current 1st-year Ph.D. students (see FORTIS Lab members).
  • Strong programming skills, demonstrated through research projects or significant open-source contributions.
  • Strong interest in open-source, and willingness to build research tools and demos.
• Prerequisites for Research Interns: Successful candidates typically have:
  • At least one top-tier publication in ML (e.g., NeurIPS, ICML, ICLR), Systems (e.g., MLSys, VLDB, OSDI), or LLMs (ACL, EMNLP, NAACL). This list is not exhaustive.
  • Strong programming skills, demonstrated through research projects or significant open-source contributions.
  • I do not have any summer interns -- I am also enjoying summer for fun :)

2. Why Join FORTIS

Compensation. Ph.D. students will receive full support outlined by the CS department (tuition waiver + stipend; currently $45,619 based on 2026-2027 funding offers). TA is always guaranteed, and we will try to give you as many RA as possible. For high-performing master/undergraduate researchers, we may consider hourly pay on a case-by-case basis. Remote intern payment remains constrained by USC policy. Please consider this before reaching out.

Compute Resources.

• fortis-prima (deployed Nov 2023): dual AMD EPYC 7763 ("Milan"), 1TB DDR4 memory (64×16GB), 15TB SSD, and 8× NVIDIA RTX 6000 Ada GPUs.
• fortis-nova (deployed Fall 2025): supports up to 8× NVIDIA RTX 6000 Blackwell Max-Q GPUs (96GB each, released in 2025); launched with 4 GPUs, with expansion to 8 GPUs planned in 2026/27.
• Cloud/API support: we provide LLM API credits, and our cloud resources primarily include AWS credits and NSF access to support large-scale experimentation.

3. How to Apply - For all positions, please fill out this Google Form: Application Form and also email the following details (subject line: "Interested in PhD/Intern at Fall 26"): Magic Words to fortis@usc.edu - please include Trojan in the email title to demonstrate that you have read these instructions carefully.

• CV: Provide a concise summary of your background and your future plans.
• Why FORTIS?: Mention any of my research papers or topics that have caught your interest. Additionally, feel free to propose new topics you are interested in exploring. I am always open to innovative ideas and fresh perspectives.