Research
Trustworthy AI for Urdu
Can AI give correct, trusted answers in Urdu, Roman Urdu and Urdu-English mixed text?
Planned research projects
All Planned. Track B comes first because it fits the BS years.
| Project | Question | Timing |
|---|---|---|
| Urdu-RomanX | Do models understand Urdu script, Roman Urdu and Urdu-English text equally well? | Years 1 to 2 |
| UrduQA-Reason | Can AI answer with evidence and multi-step reasoning, and say so when no answer exists? | Year 2 |
| UrduCodeSwitch-Bench | How reliable and safe is AI across different forms of the same language? | Years 2 to 3 |
| UrduIE-KG | Can we extract people, places, laws and relations from Urdu text? | Year 3 |
| PAKGOV-RAG-X | Can AI answer Pakistani government questions with real, checkable sources? | Years 3 to 4 |
Stretch work after these: PAK-AI-Bench (RAG reliability across English, Urdu, Roman Urdu and mixed text) and an AI public-service reliability lab. Datasets and models are released only where licenses and source permissions allow.
Research direction
One question guides my work: can AI give correct, trusted answers in Urdu, Roman Urdu and Urdu-English mixed text? I have no peer-reviewed publications yet, and I will not submit work only to reach a number.
| Working title | Timing | Status |
|---|---|---|
| Urdu-RomanX: Robust Representation Learning Across Urdu, Roman Urdu and Urdu-English Text | Draft year 2, submit year 3 | Planned |
| UrduCodeSwitch-Bench: Reliability, Factuality and Safety Evaluation Across Urdu, Roman Urdu and Code-Switched Text | Years 3 to 4 | Planned |
| PAKGOV-RAG-X: Evidence-Grounded Bilingual RAG for Pakistani Government and Public Information | Year 4 | Planned |
Research identity: ORCID 0009-0002-2163-1728
Path to an MS in AI
A plan, not a record. Dates move if grades or opportunities change, and no one can promise admission.
- Year 1, 2026 to 2027Strong grades, C++ and calculus, daily Git habit, first research reading.
- Year 2, 2027 to 2028Data structures, algorithms, machine learning basics. First Urdu NLP research project and paper draft.
- Year 3, 2028 to 2029AI core courses, code-switching benchmark, research internship applications, English test preparation.
- Year 4, 2029 to 2030PAKGOV-RAG-X as final-year project, applications for fully funded programs.
Target programs
Goals, not offers. Deadlines and rules are checked on each official page.
- MBZUAI, UAEGoal
- KAUST, Saudi ArabiaGoal
- ETH Zurich, SwitzerlandGoal
- EPFL, SwitzerlandGoal
- TU Munich, GermanyGoal