SowmithK

కర్మణ్యేవాధికారస్తే మా ఫలేషు కదాచన
మా కర్మఫలహేతుర్భూర్మా తే సంగోస్త్వకర్మణి

— Bhagavad Gītā 2.47

I am SowmithK, a 2nd year PhD student in High-Performance Automated Reasoning(HARP) Lab under Thomas Gilray. Working on accelerating logical reasoning using GPUs. Current project involves building a GPU execution engine for Datalog, building on previous work (Sun25Opt, Sun25Col, Shovon23) from the lab, to serve as GPU backend for Slog (Gilray25).

Particularly interested in building High Performance libraries in the realm of graphs and databases, building a datalog engine satisfies all three of these curiosities.

I am also working with Akash on his distributed iterative matmul to support linear recursive queries with no communication. This work is letting us focus on Linear Algebra and Relational Algebra, and the connection between them.

Always looking to chat with people, who have interests in high performance applications, databases or graphs. Please feel to reach out to my email.

news

Mar 30, 2026	Starting at relational.ai as an intern this summer, I’ll be working on their logic engine, we are still figuring out the details, looking forward to learning and doing some cool database + graphs stuff.
Nov 07, 2025	Participated in Network for Science workshop hosted by Berkeley Labs to present Akash’s poster on iterative parallel matrix exponentiation. Got to chat with Ariful Azad about CombBLAS. And see many other exciting works in the realm of graphs.
Nov 05, 2025	Completed my qualifying examination(first step towards PhD), surveyed optimal joins, and suitability for acceleration using GPUs. Focused on Acyclic Optimal Joins using Diamond Hardened Joins, and proposed ideas to simplify implementing WCOJ on GPU compared to SOTA.
May 27, 2025	Presented HARP’s published Datalog on GPU work at Minnowbrook Logic Programming Seminar, and some work in progress along with future work. The session is recorded and my talk is here.
Oct 20, 2024	Attended Programming Languages Mentoring Workshop PLMW@SPLASH24. Thank you organizers for setting up a wonderful program to help realize the breadth of PL research, and give a change to interact with Profs working on a variety of problems.

latest posts

selected publications

Datalog with First-Class Facts

Thomas Gilray, Arash Sahebolamri, Yihao Sun, and 3 more authors

Proc. VLDB Endow., Nov 2024

Abs DOI

Datalog is a popular logic programming language for deductive reasoning tasks in a wide array of applications, including business analytics, program analysis, and ontological reasoning. However, Datalog’s restriction to flat facts over atomic constants leads to challenges in working with tree-structured data, such as derivation trees or abstract syntax trees. To ameliorate Datalog’s restrictions, popular extensions of Datalog support features such as existential quantification in rule heads (Datalog*, Datalog∃) or algebraic data types (Soufflé). Unfortunately, these are imperfect solutions for reasoning over structured and recursive data types, with general existentials leading to complex implementations requiring unification, and ADTs unable to trigger rule evaluation and failing to support efficient indexing.We present DL∃!, a Datalog with first-class facts, wherein every fact is identified with a Skolem term unique to the fact. We show that this restriction offers an attractive price point for Datalogbased reasoning over tree-shaped data, demonstrating its application to databases, artificial intelligence, and programming languages. We implemented DL∃! as a system Slog, which leverages the uniqueness restriction of DL∃! to enable a communication-avoiding, massively-parallel implementation built on MPI. We show that Slog outperforms leading systems (Nemo, Vlog, RDFox, and Soufflé) on a variety of benchmarks, with the potential to scale to thousands of threads.