Detecting Manipulative Prompts in Large Language Models
An upcoming paper on detecting manipulative and adversarial prompting in large language models. Full methodology and results will be shared after publication.
Sadiq
Mansoor
I build AI that ships, not demos.
I've deployed a production RAG platform (LinkIntelligence), built a client algorithmic-trading system on Azure, and developed applied ML for signal classification and clinical prediction. I work end to end: data, models, FastAPI backends, and cloud deployment.
Work
Selected Projects
Research
Applied research
Beyond shipping products, I study how these systems actually behave, from LLM safety to clinical prediction.
Chronic Kidney Disease Prediction
Comparative machine learning for early CKD detection on the UCI clinical dataset: leakage-safe preprocessing, engineered biomarker features, and 10-fold cross-validation.
Expertise
What I do
01
LLM Systems & RAG →
End-to-end retrieval-augmented generation pipelines, prompt engineering, and LLM-powered tools. Built LinkIntelligence, a production RAG system with hybrid search and cited answers.
02
Applied ML & Research →
Signal processing, classification, and prediction models backed by real evaluation. EEG state classification, CKD prediction, and LLM safety research under review.
03
Full-Stack Web Apps →
Python and JavaScript web applications with APIs, auth, dashboards, and cloud deployment. My projects ship as working, hosted products, not just code in a repo.
Let's build something.
Looking for a software engineer, or have a project in mind? Let's talk.
Email me at sadiqmjalali111@gmail.comor send a message