Konstantinos Papakostas

Zeta Alpha

Research Engineer • Sept, 2023 — Current

Collaborating closely with enterprise clients to develop custom embedding models tailored to their proprietary data. My expertise centers on seamlessly incorporating these models into question-answering and conversational systems, boosting efficacy with domain-specific fine-tuning on the retrieval part.

Zeta Alpha

Research Intern • Nov, 2022 — July, 2023

Worked on my thesis titled Domain Adaptation for Dense Retrieval: Generating Diverse Synthetic Data with Large Language Models, under the supervision of Dr. Andrew Yates and Jakub Zavrel.

University of Amsterdam

MSc Artificial Intelligence • 2021 — 2023

Distinction: Cum Laude

Notable Classes: Deep Learning 1 & 2, Information Retrieval 1 & 2, Deep Learning for NLP
Worked as a Teaching Assistant during my second year for DL 1, FACT in AI, IR 1, and ATCS.

Aristotle University of Thessaloniki

BSc Computer Science • 2017 — 2021

Highest Grade of Class of '21

Thesis: Machine Learning for Web Robot Detection (Scalper Bots)
Minors: (1) Artificial Intelligence, (2) Data & Web Management

InRanker: Distilled Rankers for Zero-shot Information Retrieval

Applied a two-step knowledge distillation technique for cross-encoders using soft labels, achieving comparable performance between models of different sizes despite having an order of magnitude fewer parameters.

Probing LLMs for Joint Encoding of Linguistic Properties

Published in the Findings of EMNLP 2023.

Studied the information-sharing mechanisms in BERT-based models, demonstrating the presence of joint encoding across languages, linguistic entities, and tasks.

Model Analysis & Evaluation for Ambiguous QA

Published in the Findings of ACL 2023.

Investigated the modeling aspects that influence the answer quality of retrieval-augmented models in long-from ambiguous QA, along with the alignment of common evaluation metrics with human judgment.

Vision DiffMask: Faithful Interpretation of Vision Transformers with Differentiable Patch Masking

Published in the XAI4CV workshop at CVPR 2023.

Proposed an interpretation method for the Vision Transformer, which learns to mask out the subset of the input with the least impact on the output distribution, in order to analyze its decision-making process.

[Re] Exacerbating Algorithmic Bias through Fairness Attacks

Published in the ReScience Journal, presented as a poster at NeurIPS 2022.

Reproducibility study of a paper on a collection of adversarial attacks that degrade a model’s performance with regard to fairness metrics.

Leadership & team management

Served as the chairperson of a student team for 2.5 years, during which I helped coordinate various administrative subteams and organize community events on technology topics.

Communication & collaboration

Worked on a plethora of group projects as a student, many of which lead to scientific publications and/or open-source contributions.

Problem solving

As a Computer Science graduate, I've tackled many analytical problems regarding data structures and algorithms, while delivering computationally efficient solutions.

Winter School Graduate on Fairness in AI

ACM-W Greece • 2021

Participated in a collection of engaging activities, aiming to (i) understand the fairness issues in existing AI models and datasets, and (ii) propose ways to address these problems.

Seeds For The Future Graduate

Huawei • 2020

Selected for the Greek delegation for the Seeds For the Future program, during which we attended lectures and got awarded a certification on AI, 5G, IoT, and Cloud Computing.

Bronze Medal at Robotic Football

World Robot Olympiad Hellas • 2018

Worked with a small team to design a robot to play football, using LEGO MINDSTORMS EV3, and develop a custom Java framework to handle the robot's state and actions.

Best Microsoft Azure Hacks

Major League Hacking • 2018

Created a prototype website that used the municipality's open data to provide an interface for tourists to navigate through archaeological monuments, both on foot and by car.

ACM AUTh Student Chapter

Chairperson • 2019 — 2021

Coordinated an administrative team of over 25 members and helped in the organization of 15 community events about Computer Science and Career Development.