Salik Hussain

About

Biography

Short and full versions, plus the rules I work by.

Short bio

Salik Hussain is a first-year BS Computer Science student at Iqra University Islamabad who builds and evaluates retrieval systems for Urdu and Pakistani public information. His main project, PAKGOV-RAG, compares BM25, dense and hybrid retrieval on English, Urdu and mixed-language questions over official government documents. He publishes his results, including the failures, and is working toward a fully funded MS in AI.

Full biography

Salik is from Kurram District and studies in Islamabad. He finished F.Sc. Pre-Medical in August 2026 with 71.18 percent (SSC: 83.27 percent), then chose computer science after seeing how hard it is for Urdu readers to find and check government information. He began his BS in October 2026 on the 136-credit pre-medical track, with two extra mathematics courses.

Before his first class he completed Harvard CS50x, CS50 Programming with Python, CS50 AI, CS50 SQL and CS50 Cybersecurity, along with courses from IBM SkillsBuild, the University of Helsinki, Google and The Open University. He practises daily on LeetCode (795 problems solved) and holds ten HackerRank certificates.

His projects are PAKGOV-RAG, a bilingual retrieval system with a published evaluation; a change-order extraction pipeline with confidence scores and automated tests; and ChaseLog, a deployed full-stack application with database-enforced access control. Since October 2026 he is also a remote Python intern at Arch Technologies.

His research direction is trustworthy AI for low-resource languages: whether a system answers equally well in Urdu script, Roman Urdu and mixed Urdu-English text, and whether it can show the source behind each answer. By 2030 he aims to apply to MS programs in AI, including MBZUAI, KAUST, ETH Zurich, EPFL and TU Munich. These are goals, not offers, and no one can promise admission.

Working rules

  • Grades first. A strong CGPA supports everything else.
  • Finish before starting. One project at a time.
  • Be honest. No invented numbers, and no "first ever" claims without proof.
  • Respect the rules. Every dataset follows its license, privacy terms and source permissions.
  • Test everything. Results must be repeatable.
  • Targets are not facts. Goals stay goals until they are earned.