CV [pdf]

Education

Ph.D in Computer Science, UMass Lowell
M.S. in Computer Science & Engineering, Sabanci University
B.S. in Computer Engineering, Izmir Institute of Technology

Work experience

Jul 2025 – Present: Associate Director, Global AI & Advanced Analytics @ BioNTech, Massachusetts, US.
- Drive an enterprise GenAI roadmap built on reusable capabilities, balancing near-term impact with long-term platform maturity, and aligning priorities across commercial, market access, and insights needs.
- Define and lead a modular GenAI capability stack (document intelligence, knowledge grounding, content transformation, orchestration, multi-agent patterns, governance), providing hands-on technical direction for scalable implementation with LangChain (RAG/tooling), LangGraph (stateful workflows), and MCP (standardized tool/data).
- Establish a role-based GenAI operating model (e.g., interpretation, planning, gener- ation, review, validation) to standardize how GenAI is embedded into workflows and scaled across teams, including A2A collaboration patterns for specialized agent roles.
- Lead cross-functional teams across research and engineering while remaining hands- on in key technical areas (use-case scoping, evaluation design, workflow patterns) to accelerate delivery and ensure quality.
- Implement governance-by-design for regulated AI workflows, including quality bench- marks, validation loops, audit readiness, and phased delivery milestones to improve reliability and adoption.
- Lead build-vs-buy decisions and partner strategy, setting performance expectations and coordinating vendors to accelerate delivery while strengthening long-term internal ownership and scalability.
- Overall, lead the launch of 9 pharma commercial GenAI use cases and oversee on- going support for 6, driving enhancements, adoption, stakeholder feedback loops, and quality monitoring.
Apr 2025 – July 2025: Postdoctoral Researcher @ Johnson & Johnson, Massachusetts, US.
- Implemented Retrieval-Augmented Generation to ground agent reasoning in external knowledge and support iterative, feedback-driven refinement across reasoning phases.
- Designed a multi-agent generative AI system in the biomedical domain, where agents with distinct motivations (e.g., curiosity, conservatism) collaboratively explore, generate, and critique outputs derived from natural language observations.
June 2022 – August 2022: Data Science Intern @ Novartis, New Jersey, US. Novartis operates in the pharmaceutical and healthcare sector, focusing on the research, development, and commercialization of innovative medicines and healthcare solutions.
- Developed Mixed Marketing Model(s) using Turkey sales data for Jakavi and Kisqali.
- Preprocessed and mined data for the Belgium Sentiment Analysis Project.
- Reviewed literature and designed experiments for the Genomics Project on breast cancer.
Jan 2020 – Aug 2021: Sr. Machine Learning Engineer @ Cicek Sepeti, Istanbul, Turkey. Ciceksepeti.com is an online floral and gourmet foods, gift retailer operating in Turkey with subsidiaries (lolaflora.com) in other countries.
- Led a team of three junior engineers in designing and building an efficient framework to find the optimal price of products.
- Combined modern machine learning approaches with structural models to maximize company profits.
- Trained the underlying model on prior market data to predict optimal prices by learning market dynamics.
- Enabled respective employees to inspect price differences of the company’s products among e-commerce websites.
Jun 2018 – Dec 2019: Machine Learning Engineer @ Softtech, Istanbul, Turkey. SoftTech is the tech company behind IsBank, Turkey’s largest bank.
- Implemented a platform for chatbot model training and management.
- Utilizes deep learning (LSTM/GRU architectures) for larger datasets and similarity algorithms (e.g., n-grams, word mover’s distance) for smaller datasets.
- Platform is currently used by IsBank, where millions of users interact with the company’s chatbot daily.
August 2015 – June 2016: Software Engineer @ Secube, Izmir, Turkey
- Worked as a full-time software engineer during my senior year of college.
- Developed CubeBox, an application similar to GoogleDrive that allows users to store their files, access them remotely, share them with others, and encrypt files before storing them, unlike other available apps.
- Developed a mobile application for members of the Turkish parliament to securely sign and share documents.

Research experience

August 2021 – Mar 2025: Research Assistant @ Persistent Autonomy and Robot Learning lab (PeARL), University of Massachusetts Lowell.
- Currently, exploring how Large Language Models (LLMs) can be used within Multi-Agent Reinforcement Learning (MARL) frameworks to improve agents’ performance and encourage the emergence of novel behaviors.
- Conducted research on MARL and emergent behaviors in collaborative environments. Designed and implemented a graph-based approach to guide and influence the agents’ behavior in multi-agent settings.
- Implemented existing single-agent deep reinforcement learning architectures (DQN, A2C, DDPG, PPO). Implemented and extended existing multi-agent deep reinforcement learning architectures, both value-based (VDN, QMix) and policy-based (MADDPG, MAPPO).
- Designed and implemented a novel algorithm, D-CBRS, to address intra-class diversity in continuous learning.
September 2016 – July 2021: Research Assistant @ Behavioral Analytics and Visualization Lab (BAVLAB), Sabanci University. BAVLAB is cofounded by Sabanci University and MIT Media Lab Human Dynamics Group
- Conducted research on human-agent interaction, integrating methods such as visualization, computer vision, and natural language processing.
- Developed an analytical agent capable of interpreting the environment and generating insights for data analysis based on presented data and cues from human subjects.
- Performed research on collaborative computer-aided designs using a tabletop setup.
- Conducted studies on Twitter data and analyzed patterns using data mining methods.
September 2015 – May 2016: Research Student @ Visual Intelligence Research Group, Izmir Institute of Technology.
- Worked on obstruction removal from images and implemented a computational approach to remove impediments such as fences, windows and reflections from images.

Technical skills

AI/ML & Analytics: Deep Learning, Transformers, GenAI, Retrieval-Augmented Generation, Reinforcement Learning, Multi-agent Systems, NLP, A/B testing, Forecasting, Statistical Modelling, Predictive Analysis, Tableau, PowerBI
Languages: Python, C, Matlab, R, Object-Oriented (Java/C++/C#), Web Development (HTML/CSS/Javascript/PHP)
Frameworks: PyTorch, Tensorflow, LangChain, LangGraph, MCP, A2A, Keras, Flask, OpenCV, OpenGL, D3, Android
Others: Databricks, AWS, GCP, Docker, Neo4j, Kafka, Solr, MongoDB, Git

Publications

Collaborative Adaptation for Recovery from Unforeseen Malfunctions in Discrete and Continuous MARL Domains

Y. Findik, H. Hasenfus, R. Azadeh. "Collaborative Adaptation for Recovery from Unforeseen Malfunctions in Discrete and Continuous MARL Domains." In proceeding 63rd IEEE Conference on Decision and Control (CDC), Milan, Italy, pp. xx-xx, Dec. 16-19, 2024.

Relational Q-Functionals: Multi-Agent Learning to Recover from Unforeseen Robot Malfunctions in Continuous Action Domains

Y. Findik, P. Robinette, K. Jerath, R. Azadeh. "Relational Q-Functionals: Multi-Agent Learning to Recover from Unforeseen Robot Malfunctions in Continuous Action Domains." In proceeding 21st IEEE International Conference on Ubiquitous Robots (UR), New York, USA, pp. 251--256, June 24-27, 2024.

Mixed Q-Functionals: Advancing Value-Based Methods in Cooperative MARL with Continuous Action Domains

Y. Findik, R. Azadeh. "Mixed Q-Functionals: Advancing Value-Based Methods in Cooperative MARL with Continuous Action Domains." arXiv preprint arXiv:2402.07752, February 2024.

Impact of Relational Networks in Multi-Agent Learning: A Value-Based Factorization View

Y. Findik, P. Robinette, K. Jerath, R. Azadeh. "Impact of Relational Networks in Multi-Agent Learning: A Value-Based Factorization View." In proceeding 62nd IEEE Conference on Decision and Control (CDC), Marina Bay Sands, Singapore, pp. 4447-4454, Dec. 13-15, 2023.

Influence of Team Interactions on Multi-Robot Cooperation: A Relational Network Perspective

Y. Findik, H. Osooli, P. Robinette, K. Jerath, R. Azadeh. "Influence of Team Interactions on Multi-Robot Cooperation: A Relational Network Perspective." In proceeding IEEE International Symposium on Multi-Robot and Multi-Agent Systems (MRS), Boston, MA, Dec 4-5, 2023.

Collaborative Adaptation: Learning to Recover from Unforeseen Malfunctions in Multi-Robot Teams

Y. Findik, P. Robinette, K. Jerath, R. Azadeh. "Collaborative Adaptation: Learning to Recover from Unforeseen Malfunctions in Multi-Robot Teams." MADGames workshop at IEEE/RSJ International Conference on Intelligent Robots and Systems, Detroit, MI, USA, Oct 1-5, 2023.

D-CBRS: Accounting for Intra-Class Diversity in Continual Learning

Y. Findik, F. Pourkamali-Anaraki. "D-CBRS: Accounting for Intra-class Diversity in Continual Learning." The 29th IEEE International Conference on Image Processing (ICIP), October 2022.

Facilitating Decision-Making with Multimodal Interfaces in Collaborative Analytical Sessions

Y. Findik, H.A. Boz, B. Bozkaya, S. Balcisoy. "Facilitating Decision Making with Multimodal Interfaces in Collaborative Analytical Sessions." In Textbook of Intelligent Scene Modeling and Human-Computer Interaction. Human-Computer Interaction Series. Springer, Cham, June 2021.

Question Similarity Detection in Turkish Using Semantic Textual Similarity Methods

E. Yildiz and Y. Findik. "Question Similarity Detection in Turkish Using Semantic Textual Similarity Methods." In proceeding 27th IEEE Signal Processing and Communications Applications Conference, April 2019.

Twitter Reveals: Using Twitter Analytics to Predict Public Protests

M. Bahrami, Y. Findik, B. Bozkaya, S. Balcisoy. "Twitter Reveals: Using Twitter Analytics to Predict Public Protests." arXiv preprint arXiv:1805.00358, May 2018.

Low-fidelity Prototyping with Simple Collaborative Tabletop Computer-aided Design Systems

E. Kaya, S. Alacam, Y. Findik, S. Balcisoy. "Low-fidelity Prototyping with Simple Collaborative Tabletop Computer-aided Design Systems." Computers & Graphics, 70 , February 2018.

Talks

Relational Q-Functionals (UR’24)

June 25, 2024

Paper presentation at New York University, New York NY, USA

Designing Algorithms and Systems for Multi-agent Collaboration

December 12, 2023

Guest Lecture at University of Massachusetts Lowell, Lowell MA, USA

Influence of Team Interactions on Multi-Robot Cooperation (MRS’23)

December 05, 2023

Paper presentation at Boston University, Boston MA, USA

Collaborative Adaptation Framework (IROS’23)

October 01, 2023

MADGames@IROS talk at Huntington Place, Detroit MI, USA

D-CBRS: Accounting For Intra-Class Diversity in Continual Learning (ICIP’22)

October 16, 2022

Poster presentation at Palais 2 l'Atlantique, Bordeaux, France

Question Similarity Detection in Turkish Using Semantic Textual Similarity Methods

April 25, 2019

Paper presentation at Sivas Cumhuriyet University, Sivas, Turkey

Teaching Experience

Introduction to Data Science

Served as a Teaching Assistant for the Introduction to Data Science course. My responsibilities included leading lab sessions for three hours per week and supervising the projects of 40 students. Additionally, I graded progress reports and code submissions, and provided each group with feedback to guide their subsequent project steps.

Computer Graphics

As a Teaching Assistant for the Computer Graphics course, I held weekly lab sessions, 2 hours per week. I supervised and provided guidance for 50 student projects, evaluated progress reports and code submissions, and offered detailed feedback to each group to facilitate further project development.

Introduction to Computing

Served as a Teaching Assistant for the Introduction to Computing course over five semesters. Responsibilities included conducting weekly lab sessions for three hours, and grading quizzes and exams.

Yasin Findik