Matan Shtepel
CV
 
                Howdy! I'm a second-year PhD student at Carnegie Mellon working on AI Safety with the epic Andrew Ilyas and partially funded by NSF GRFP. 
                    Office: Gates 9213.
                    
                    
                    Previously at CMU, UPenn, and UCLA I worked on cryptography (including zkSNARKs, MPC, PIR, and ORAM). I switched to work on AI Safety because I think it is super interesting and extremely important! I help out with CASI.
                    
                    Outside of research and building subpar academic websites, I also enjoy 
                    fashion, Tex Crick, running, reading, beans, and collecting magazines!
                    
                    
                    
                    
                    
                    Contact me for any reason! Especially happy to hear from (ex) community college students interested in research! matan.shtepel@gmail.com
                        
                
"It is by logic that we prove, but by intuition that we discover." (Poincaré)
Preprints
- 
                        Query Optimal IOPPs for Linear Time Encodable Codes   
 A. Baweja, P. Mishra, T. Mopur (UPenn), M. Shtepel (CMU)
 We give IOPPs for linear-time encodable codes with provably optimal query-complexity.
- 
                        FICS and FACS: Fast IOPPs and Accumulation via Code-Switching 
                          
 A. Baweja, P. Mishra, T. Mopuri (UPenn), M. Shtepel (CMU)
 We give IOPPs for linear-time encodable codes with state-of-the-art query-complexity.
Publications
- 
                        Malicious Security for PIR (almost) for Free
                         (CRYPTO '25)  
 B. Falk, P. Mishra, M. Shtepel (UPenn)
 We describe a construction of malicious PIR from locally decodable codes with very low overhead.
- 
                        DORAM revisited: Maliciously secure RAM-MPC with logarithmic overhead (TCC '23) 
 B. Falk, D. Noble (UPenn), R. Ostrovsky, M. Shtepel, J. Zhang (UCLA)
 We give the state-of-the-art construction of maliciously-secure Distributed ORAM.
- 
                        
                        GigaDORAM: Breaking the Billion Address Barrier (USENIX '23) 
 B. Falk (UPenn), R. Ostrovsky, M. Shtepel, J. Zhang (UCLA)
 We show devise and implement a Distributed ORAM construction orders of magnitude more efficient than prior work. Code, Talk.
-  "On Totalization of Computable Functions in a Distributive
                            Environment" ,(International Journal of Parallel, Emergent and
                        Distributed Systems (GPAA))
 M. Burgin, M.Shtepel (UCLA)
 We investigate which classes of computations can be totalized, i.e. modified so every input maps to an output in a `consistent' way
In alphabetical order.
Selected Photos
 
                        Curly boys present an early version of Inspect WandB @ MARS 3.0
 
                        Friendsgiving at Gaurav and Barry's, (Friendship, Pitt).
 
                        Stereotypical Theory group outing = brewery 🍻
Academic Aspiration
As an (aspiring) academic, I feel obligated to say there will be a blog coming soon. Don't worry! I will update twice during summer, then never again after 😏. If it will ever exist, it will probably be here.
 
                         
                         
                         
                         
                         
                         
                         
                        