Industry
AI assistants, APIs, and mobile applications. My work includes code-generation tools and message classification.
Browse projectsI’m an AI engineer and master’s student in Computer Engineering. I build applications that use language models, and I study how those models reason and where they fail.
Before joining Gachon University, I worked on AI assistants at AIO and iOS applications at FitFlex. My experience covers research, backend development, and mobile software.
AI assistants, APIs, and mobile applications. My work includes code-generation tools and message classification.
Browse projectsExperiments on language-model reasoning and safety, including when models give incorrect answers or bypass safeguards.
Read the papersTools for querying restaurant data, analyzing customer reviews, and turning spoken information into structured offers.
View applicationsGachon University · Seongnam, South Korea
Research: language-model security and reasoning. Coursework includes reinforcement learning, natural language processing, and computer vision.Air University · Islamabad, Pakistan
Final-year project: PTSD prediction from clinical text, with mentorship through the 10Pearls FYP Accelerator.Served as a reviewer for Women in Machine Learning (WiML) at NeurIPS 2026.
Presented When Reasoning Collapses as a poster in Seoul, South Korea.
Presented work on language-model reasoning and jailbreak robustness.
Reviewed applications for the conference selection process.
Participated in sessions on responsible AI, language models, and federated learning at QCRI in Doha.
Contribute to government-funded work on language-model security and trustworthy AI at Gachon University.
How models plan, use tools, and maintain reliable reasoning as tasks grow more complex.
Adversarial robustness, jailbreak evaluation, differential privacy, and secure ML pipelines.
Diffusion models, image–text systems, and the ways multiple modalities can work together.
Energy-efficient models, carbon-aware training, and practical AI under resource constraints.
Applications in green finance, ESG analysis, and climate-resilient financial systems.
AI, topic by topic: articles, courses, notes, and related YouTube and LinkedIn posts.
More details to come.
I study reasoning failures and security weaknesses in large language models.
I built AI assistants and APIs for business data, customer feedback, and healthcare-related applications.
I developed Swift applications and added machine learning to mobile products.
Python / audio forensics intern
Machine learning tutor
FYP Accelerator fellow
Graduate trainee, applied AI
Innovation consultant intern
M Labs fellow
GitHub updated 2026-11-02
Includes projects and learning collections. Duplicate portfolio sites and empty placeholders are omitted.
I study the reasoning and safety of large language models at ISML Lab. My work uses experiments to examine when models make mistakes and how their behavior changes under more difficult tasks or adversarial prompts.
Testing how reasoning performance changes as tasks require more inference steps, with direct-answer and reasoning-prompt comparisons.
Compared direct-answer and reasoning prompts across CLUTRR and ProofWriter using five language models. The evaluation measured accuracy at increasing reasoning depths; improvements at shallow depths often weakened or reversed as tasks became more complex.
A tri-agent reinforcement learning framework for automated adversarial testing and safety evaluation of language models.
Used a three-agent reinforcement learning framework to automate adversarial prompting and evaluate whether language models could be induced to bypass safeguards.
Investigating when additional reasoning helps factual question answering, and when selecting better evidence matters more.
Examines the separate contributions of reasoning and evidence selection to factual question answering. Accepted-poster status is taken from my resume; no code repository was linked there.
I’m the founder of an AI business offering products and project-based services, including custom development and AI consulting.
More details to come.
I’m a cofounder of an agritech product that we plan to develop into a business.
Product in development. Business details to come.
For questions about my work, research collaboration, or an AI development project, you can reach me here.
azkaikramullah496@gmail.comExperience, education, publications, and projects in one PDF.