Nikolai Karpov

Email: kimaska@gmail.com

Research Engineer at Logical Intelligence, developing efficient pipelines for fine-tuning core LLMs. PhD in Computer Science from Indiana University, Bloomington.

Previously at Oxford Nanopore Technologies, core developer of chinook for CYP2D6 analysis.

20+ publications (282 citations, h-index 9) in top-tier venues: NeurIPS, AAAI, FOCS, VLDB, PODS.

Author of editdistancek Rust library (200K+ downloads). Contributor to whatshap genomic phasing tool.

Awards: Paul Purdom Fellowship, Research Assistant of the Year (2022), 28th place ACM ICPC World Finals (2015), Microsoft Fellowship nominee.

Sometimes I participate in programming contests (meta profile).

Conference papers

Nearly tight bounds for exploration in streaming multi-armed bandits with known optimality gap

AAAI 2025

Parallel Best Arm Identification in Heterogeneous Environments

SPAA 2024

Communication-Efficient Collaborative Regret Minimization in Multi-Armed Bandits

AAAI 2024 [eprint]

MinJoin++: A Fast Algorithm for String Similarity Joins under Edit Distance

VLDB Journal

Communication-Efficient Collaborative Best Arm Identification

AAAI 2023 [eprint]

SyncSignature: A Simple, Efficient, Parallelizable Framework for Tree Similarity Joins

VLDB 2023

Instance-Sensitive Algorithms for Pure Exploration in Multinomial Logit Bandit

AAAI 2022 [eprint]

Batched Coarse Ranking in Multi-Armed Bandits

NeurIPS 2020 [eprint]

Collaborative Top Distribution Identifications with Limited Interaction

FOCS 2020 [eprint]

Distributed and Streaming Linear Programming in Low Dimensions

PODS 2019 [eprint]

Sketching Algorithms for Genomic Data Analysis and Querying in a Secure Enclave

RECOMB 2019 [eprint]

A Multi-labeled Tree Edit Distance for Comparing "Clonal Trees" of Tumor Progression

WABI 2018 [eprint]

An exponential lower bound for cut sparsifiers in planar graphs

IPEC 2017 [eprint]

Parameterized Algorithms for Partitioning Graphs into Highly Connected Clusters

MFCS 2017 [eprint]

Parameterized Complexity of Secluded Connectivity Problems

FSTTCS 2015 [eprint]

Parameterized Complexity of Superstring Problems

CPM 2015 [eprint]

Manuscripts

Batched Thompson Sampling for Multi-Armed Bandits