vita
Basics
| Name | Ishrak Hayet |
| Label | CS PhD Student @ NC State University |
| ihayet@ncsu.edu | |
| Url | https://ihayet.github.io/ |
| Summary | Fourth-year CS PhD Student at North Carolina State University. Advised by Prof. Marcelo d'Amorim. Current research focus is on applying ML/AI/LLM techniques for code/test generation/completion, building benchmarks, and rigorous empirical evaluation of agentic systems using industry-standard metrics. Earned CS MS degree from the University of Kansas under the co-supervision of Dr. Bo Luo and Dr. Zijun Yao. Former member of the KU InfoSec lab (now known as HASSC lab). The research focus was to find a threat model related to the training data privacy of fine-tuned language models. Passionate about and experienced in innovative applications of Machine Learning and Artificial Intelligence across a wide range of application areas, including but not limited to code generation, code completion, privacy exposure, brain signal classification, and text completion. Experienced with data collection, cleaning, and curation for training, fine-tuning, and evaluating large-scale optimization problems. Have strong foundations of machine learning, deep learning, statistical models, supervised and unsupervised learning methods, and hypothesis testing. Proficient in object-oriented programming in Python and Java. Extremely efficient in collaborating with generative AI models and agents with manual validation for solving everyday tasks 10x faster. |
Work
-
2022.08 - Present Graduate Student Researcher
Software Engineering Research Lab, NC State University
1. Collecting and preprocessing custom datasets. 2. Training/fine-tuning ML/AI/LLM models using curated datasets for specialized downstream tasks. 3. Rigorously and empirically evaluating model/system outputs using industry-standard metrics. 4. Built a code completion technique using feedback-directed mock generation and neural type inference, achieving 80 % code coverage. 5. Developed a prompt engineering framework to generate and repair JUnit test oracles using LLMs, beating SoTA accuracy by more than 10 %.
-
2019.08 - 2022.05 Graduate Research and Teaching Assistant
Information Assurance Research Lab, University of Kansas
1. Built a deep neural model to infer private natural language datasets from pre-trained and fine-tuned embeddings with 70 % accuracy. 2. Trained Word2Vec, GloVe, and BERT vectors and built LSTM, CNN, and Transformer models for datasets with 20K samples. 3. Conducted research on information security and privacy of natural language embedding models. 4. Accepted Paul F. Huebner Memorial Award for becoming an outstanding GTA. 5. Designed and conducted labs on Operating Systems topics using C on Linux.
-
2018.01 - 2019.07 Software Developer
Porate Chai EdTech
1. Implemented a matchmaking Android app with a RESTful Node.js application server; accessed by 50K users 2. Designed application system, matchmaking server, and used local SQLite database caching to optimize loading times to less than 1ms
Education
-
2022.08 - Present Raleigh, North Carolina
-
2019.08 - 2022.05 Lawrence, Kansas
-
2014.01 - 2017.12 Dhaka, Bangladesh
Awards
- 2024.09.01
Travel Grant
ACM SIGSOFT
This award partially supported my travel to present our paper at ACM ISSTA 2024 in Vienna, Austria.
- 2022.03.01
Distinguished Teaching Assistant
University of Kansas (KU)
I was recognized as a distinguished Graduate Teaching Assistant for my contributions to teaching duties at KU.
- 2021.11.08
Registration Scholarship
Asian Conference on Machine Learning (ACML)
I was awarded free registration to attend the Online Asia Machine Learning School (OAMLS) conference.
- 2014.01.01
OIC Scholarship
Organization of Islamic Cooperation
This international scholarship fully covered my undergraduate tuition, lodging, and monthly stipends.
- 2009.04.01
Exchange Scholarship
ECA, US Department of State
I was awarded this scholarship to conduct high school STEM research on a Bangladeshi river and to present our findings at the YouthCaN 2009 conference in New York City.
Certificates
| Attention Models | ||
| DeepLearning.AI | 2022-11 |
| Fine-tuning BERT | ||
| DeepLearning.AI | 2022-06 |
| Vector Spaces | ||
| DeepLearning.AI | 2020-08 |
| SDE Fundamentals | ||
| Microsoft | 2014-12 |
Publications
-
2025 ChatAssert: LLM-based Test Oracle Generation with External Tools Assistance
IEEE TSE
This paper proposes ChatAssert, a prompt engineering framework that can generate jUnit test oracles using large language models. ChatAssert combines a generation phase with summaries and few-shot examples, and a repair phase with static and dynamic oracle repair.
-
2024 Incompleter: Feedback-directed Partial Execution
ACM ISSTA
In this paper, we propose Incompleter, a partial code execution technique that inductively mocks undefined code elements based on error feedback and eventually unmocks the mocked objects using a deductive and neural type inference. Our evaluations showed that bugs can still be found from partial code snippets if they are executed with Incompleter.
-
2022 Invernet: An Inversion Attack Framework to Infer Fine-tuning Datasets through Word Embeddings
EMNLP Findings
We propose Invernet as an inversion model that can infer word-to-word co-occurrence statistics using the differential embeddings between a pretrained and a fine-tuned language model. We also showed that simple decoding approaches can be used to fairly reconstruct private training dataset using the inferred word-to-word co-occurrence statistics.
-
2019 Designing a Hierarchical Keyboard Layout for Brain Computer Interface based Text Entry
IEEE International Conference on Electrical, Computer, and Communication Engineering
In this paper, we propose an accessiblity-aware text entry technique with an EEG-based input modality. Our technique combines a hierarchical motor signal classifier and a hidden markov model based text completion module that are combined in a soft keyboard layout.
Languages
| English | |
| Fluent |
| Bangla | |
| Native |
| Hindi | |
| Conversational |
Interests
| Agent Development |
| Machine Learning |
| Large Language Models |
| Natural Language Processing |
| AI Engineering |
Volunteer
-
2017.06 - 2017.12 Dhaka, Bangladesh
Head of Mobile App Dev Contest
Computer Society, IUT
I was responsible for organizing and coordinating the activities for mobile app development contest at one of the largest national ICT competitions held at IUT.
-
2017.01 - 2017.12 Dhaka, Bangladesh
Head of Software Systems
Mars Rover Project, IUT
I led the software team for the IUT Mars Rover project which ran successful campaigns at various international rover competitions. Our team developed the navigation, motor control, and communication software for a life-sized mars rover robot.
-
2008.01 - 2010.12 Dhaka, Bangladesh
Tutor
St. Joseph Higher Secondary School
Teaching various subjects to underprivileged children after regular school hours.