Hello, I am Salik Hussain, a Computer Science student from Rawalpindi, Pakistan.

I started my BS Computer Science at Iqra University Islamabad in October 2026. My school background is Pre-Medical (F.Sc. PCB), so I follow the 136-credit pre-medical track with two extra math courses. Before my first class, I finished 10 online certifications in programming, AI, databases and security (see below).

My research question in one line:

Can AI give correct, trusted answers in Urdu, Roman Urdu and Urdu-English mixed text?

I am building PAKGOV-RAG, a bilingual Urdu-English retrieval-augmented generation (RAG) system for Pakistani government documents. Its current stack is Python, Git, FAISS and LangChain. I am still learning PyTorch and Hugging Face.

My promise: everything on this page is either finished, in progress, or clearly marked Planned. I do not claim results I have not earned yet.

Certifications · Big Picture · Priorities · Year 1 · Year 2 · Year 3 · Year 4 · Skill Projects · Research · Papers · Internships · Letters · Products · Clubs · Tests · MS Targets · Skills

All 10 were issued in August and September 2026, before my university classes began. Certificates show learning. They are not research results, and I list them honestly as that.

pie showData
    title Certifications by Provider
    "Harvard CS50" : 4
    "IBM SkillsBuild" : 2
    "Google" : 1
    "The Open University" : 1
    "University of Helsinki" : 1
    "Anthropic" : 1

mindmap
  root((Salik Hussain))
    Degree
      Iqra BS CS
      136 credits
      CGPA 3.85 plus
    Done Already
      10 certifications
      PAKGOV-RAG v0
    10 Skill Projects
      C++ to Python
      Databases and Search
      Final Year AI Product
    10 Research Projects
      Urdu NLP track
      Reliability and Products track
    6 Papers
      3 core
      3 stretch
    6 Internships
      LUMS
      NUST
      Abroad funded
      Remote
      Industry
    6 Letters
    Products
      AI Reliability Lab
      GovAI Studio
      UrduAI Enterprise
      RegulaPak
      PakVoice
    Tests
      NCT
      IELTS or TOEFL
      GRE if needed
    Final Goal
      Funded MS in AI

gantt
    title Master Plan 2026 to 2030
    dateFormat YYYY-MM
    axisFormat %b %Y
    section Done
    10 certifications              :done, 2026-08, 2026-09
    section Degree
    Year 1 Sem 1-2                 :active, 2026-10, 2027-07
    Year 2 Sem 3-4                 :2027-10, 2028-07
    Year 3 Sem 5-6                 :2028-10, 2029-07
    Year 4 Sem 7-8                 :2029-10, 2030-07
    section Core Research
    PAKGOV-RAG v0 upgrade          :2026-10, 2027-06
    Urdu-RomanX                    :2027-01, 2027-12
    UrduQA-Reason                  :2027-10, 2028-07
    UrduCodeSwitch-Bench           :2028-01, 2029-03
    UrduIE-KG                      :2028-10, 2029-07
    PAKGOV-RAG-X                   :2028-10, 2030-03
    section Papers
    Paper 1 Urdu-RomanX            :2027-10, 2028-08
    Paper 2 CodeSwitch-Bench       :2028-10, 2029-08
    Paper 3 PAKGOV-RAG-X           :2029-06, 2030-03
    section Internships
    NUST SEECS                     :2027-12, 2028-02
    LUMS Dr. Agha                  :2028-06, 2028-08
    Remote research                :2028-12, 2029-02
    Funded abroad                  :2029-06, 2029-08
    Industry AI or SE              :2029-10, 2030-01
    section Applications
    English test and shortlist     :2029-06, 2029-09
    SOP, CV and letters            :2029-08, 2029-11
    Submit MS applications         :2029-10, 2030-01
    Interviews and decisions       :2030-01, 2030-05

pie showData
    title Credit Hours by Category
    "General Education" : 34
    "CS Core" : 48
    "Specialization Electives" : 24
    "Allied / IDS Math" : 12
    "Math Deficiency" : 6
    "Final Year Project" : 6
    "Internship" : 3
    "Professional Certification" : 3

🥇 CGPA → 🥈 AI/CS skills → 🥉 Research and projects → 4️⃣ Internships → 5️⃣ Clubs

Grade plan: A or A+ in programming, math, algorithms and AI. B+ is acceptable in general education. Weekly routine: 5 short math sessions + 4 programming sessions + 1 review session. I keep an error notebook and review it weekly. Rule: I finish one project at a time. I never start five big projects together.

The curriculum follows the Iqra 136-credit pre-medical plan. Items marked ⚠️ must be confirmed with the department (see the question list).

Goal: Strong grades, solid C++, algebra and calculus, Git habit, first research reading. Head start: My CS50x, Open University C, CS50 SQL and CS50 AI certificates already cover C, SQL and AI basics. C++ in Semester 1 is the new part.

Self-study: C++ on learncpp.com · HTML and CSS · Git and GitHub · 20 to 30 minutes of math daily · LeetCode practice. Project 1: Student Marks Calculator (C++). Target: Explain and write small C++ programs without copying, and solve foundation math alone.

Self-study: Python projects · a C++ library-management project. Project 2: Portfolio Website (HTML, CSS, GitHub). Project 3: Student Record System (OOP and file handling). Research prep: Learn to read an abstract, find the research question, and write a one-page literature summary. Review PAKGOV-RAG Version 0 and list upgrades. Plan Urdu-RomanX. Do not rush to publish.

• 10 certifications finished before university
• CGPA 3.85 or higher
• Projects 1, 2, 3 on GitHub with clear READMEs
• Daily or near-daily small commits
• PAKGOV-RAG v0 reviewed
• Know my professors in programming and math (future letters)

Goal: Master data structures and algorithms, learn ML basics, finish my first research project and first paper draft.

Self-study: NumPy · pandas · basic Linux · Git branches and pull requests. Project 4: Data Structures Visualizer. Project 5 (starts): Student Performance Analytics (SQL and Python). Research: Read 10 papers in Urdu NLP or multilingual retrieval. Make a table: problem, method, dataset, results, limits.

Self-study: Linux · algorithm complexity · testing · probability and statistics · scikit-learn · matplotlib. Project 6 (starts): Urdu-English Text Classifier. Research: Write the question, dataset, baseline, metrics and limits before running experiments.

• CGPA 3.85 or higher
• Projects 4, 5, 6 done
• Urdu-RomanX results and Paper 1 draft
• First research supervisor
• CV reviewed, no "random certificate" list

Goal: Build the benchmark, win a funded abroad research place, finish Paper 2, start my first product.

Self-study: scikit-learn · PyTorch (tensors, autograd, training loops) · train/validation/test splits · baseline vs complex model. Project 6 (finish): Urdu-English Text Classifier. Project 7: Urdu Document Search Engine. Apply in autumn 2028 to the funded abroad programs (see Internships).

⚠️ The elective list is a proposal. I will ask for the official list and choose the closest match. Self-study: transformers · embeddings · a bilingual Urdu-English retrieval project. Project 8 (starts): PAKGOV-RAG improvement (retrieval metrics, source grounding). Research milestone: Submit a poster, workshop paper or student research proposal only if the work is genuinely new and the venue is credible.

• CGPA 3.85 or higher
• Applied to 4 abroad programs in autumn 2028
• Paper 2 ready, one serious research project in progress
• Projects 7 and 8 started or done
• English test booked

Goal: Change from "good undergraduate" to "competitive international MS applicant".

Self-study: Reproduce one published paper · pick my supervisor. Project 10 (starts): Final-Year AI Product (authentication, logs, tests, deployment guide).

Final deliverables: tested code, documentation, permitted data, methodology, baselines, results, limits, ethics, demo and final report.

• CGPA does not collapse in final year (3.85 or higher)
• FYP finished and published as open source
• SOP, CV and letters done by October 2029
• Applications sent before each official deadline
• Funding and scholarship forms filed

I finish fewer projects well rather than publishing many half-done repositories.

Every project gets a README, install steps, data sources, limits, tests, examples, and a clear note of what I built myself. For government or legal documents I keep source links and document IDs, and I never claim legal accuracy.

Two tracks, one coherent field: trustworthy AI for Urdu and low-resource languages.

Order rule: Track B (Urdu NLP) comes first because it matches my years. Track A projects overlap with it (for example PAK-AI-Bench shares code with PAKGOV-RAG-X), so I merge them where possible. A2, A4 and A5 are stretch goals that may continue after the BS or during the MS.

My current PAKGOV-RAG is Version 0 (pilot). I upgrade it step by step.

flowchart LR
    A[Official documents] --> B[Cleaning and chunking]
    B --> C[BM25 search]
    B --> D[Dense search]
    C --> E[Hybrid retrieval]
    D --> E
    E --> F[Reranker]
    F --> G[LLM answer]
    G --> H[Citation check]
    H --> I[Answer with sources]

Measured with: Recall@5, Recall@10, MRR, nDCG, answer correctness, groundedness, citation precision and recall, unsupported claims, hallucination, abstention, speed and cost.

⚠️ Every dataset must respect licenses, privacy, platform terms and source permissions.

Honest rules: A paper is an outcome, not a guarantee. The real contribution decides the venue. I never submit just to reach a number. All six are Planned.

timeline
    title Internship Slots
    Winter 2027-28 : NUST SEECS research
    Summer 2028 : LUMS with Dr. Agha Ali Raza
    Winter 2028-29 : Remote research
    Summer 2029 : Funded abroad program
    Fall 2029 : Industry AI or software
    Backup : LUMS Summer Research Programme

Abroad programs (apply in autumn 2028, apply to all):

I will check dates and eligibility on every official page. I have not confirmed the 2028-29 cycle. Contacting a supervisor is a request, not a confirmed vacancy. If one slot fails, I use the backup.

Most MS programs ask for about 3 letters. I build 6 relationships and send the best 3 for each application. A letter is never guaranteed. I aim for people who directly supervised my work and can describe my real contribution.

When to ask: Introduce myself to professors from Year 1. Ask for letters in September 2029, with 4 to 6 weeks of notice. Send each writer my CV, SOP draft, transcript, project links and a deadline list.

I merge everything into one company idea, PAKAI. I will not invent prices, users or revenue before I have real customers.

flowchart TD
    R[Research and Benchmarks] --> P1[AI Reliability Lab<br/>also called PAKAI Verify]
    R --> P2[GovAI Studio]
    R --> P3[UrduAI Enterprise]
    R --> P4[RegulaPak]
    R --> P5[PakVoice]
    P1 --> U[Real users and feedback]
    P2 --> U
    P3 --> U
    P4 --> U
    P5 --> U
    U --> C[More research credibility]

UrduAI Enterprise APIs (build only what users need): /search · /qa · /summarize · /extract · /classify · /rag · /evaluate

APIs: PAK-RAG API (search, retrieval, evidence) · PAKAI Verify API (evaluate an AI system) · PakVoice API (speech → AI → evidence-grounded answer).

RAG development · AI evaluation reports · AI automation · AI dashboards · multilingual AI testing · research engineering (only after real proof of skill) · AI training. Freelancing starts small in Year 1 to 2 and shrinks when research and internships matter more.

GitHub rule: a few serious repositories beat many empty ones. Planned names: PAKGOV-RAG · PAK-AI-Bench · PAK-Multilingual-AI · AI-Public-Service-Reliability-Lab · RegulaPak · PakVoice · PAKAI-Verify.

Maximum 5 clubs plus 2 offices. Grades always come first.

gantt
    title Clubs and Offices by Semester
    dateFormat YYYY-MM
    axisFormat %b %y
    section Clubs
    Softech                :2026-10, 2030-06
    Cyber Security Force   :2027-02, 2029-07
    Innovation Club        :2027-08, 2030-01
    Freelance Force        :2027-02, 2028-07
    Debate Club            :2026-10, 2028-07
    section Offices
    International Office   :2026-10, 2030-06
    Career Services        :2027-02, 2030-06

Question for the International Office (Sem 1): "I am a BS CS student planning for an MS in CS/AI/ML abroad. What semester-exchange and student-mobility opportunities are available to BS students, and what are their official eligibility requirements and deadlines?" I also ask about partner universities, CGPA rule, credit transfer, funding and English requirements.

gantt
    title MS Application Calendar
    dateFormat YYYY-MM
    axisFormat %b %Y
    section Preparation
    Shortlist universities and professors  :2029-06, 2029-09
    English test IELTS or TOEFL            :2029-06, 2029-09
    GRE only if required                   :2029-07, 2029-10
    section Documents
    SOP drafts                             :2029-08, 2029-11
    Final CV                               :2029-09, 2029-10
    Ask letter writers                     :2029-09, 2029-10
    Request transcripts                    :2029-09, 2029-11
    section Apply
    Submit applications                    :2029-10, 2030-01
    Funding and scholarship forms          :2029-10, 2030-01
    section After
    Interviews and decisions               :2030-01, 2030-05
    Visa and travel                        :2030-05, 2030-08

Exact deadlines differ for every university. I use each official page and never rely on this chart alone.

Degree names to apply for: MS Computer Science · MS Artificial Intelligence · MS Machine Learning. Funding: I apply to fully funded options and scholarships. No one can promise admission. It depends on grades, research fit, letters and funding.

Computer Science researcher focused on low-resource multilingual AI, trustworthy LLM evaluation and evidence-grounded RAG, with open benchmarks, reproducible systems, research papers and AI products for multilingual public-information applications.

1. The names and codes of both math deficiency courses
2. Which semester Calculus and Linear Algebra fall in for my cohort
3. The official elective list and prerequisites
4. Which Quantitative Reasoning and IDS-III/IV courses include Probability & Statistics
5. Whether PHP is graded in the ICT lab
6. Whether a research or remote internship can count for CMC493

Tools marked "learning" are skills I am still building.

Learning cycle: understand simply → worked examples → solve alone → record errors → revisit after a few days. The goal is mastery, not finishing video playlists.

learncpp.com · Harvard CS50 · Hugging Face Learn · Kaggle Learn · ACL Anthology · arXiv cs.CL · Papers with Code

1. Grades first. A strong CGPA supports everything else.
2. Finish before starting. One project at a time.
3. Be honest. No fake numbers and no "first ever" claims without proof.
4. Respect rules. Every dataset follows licenses, privacy and permissions.
5. Test everything. Results must be repeatable.
6. Targets are not facts. Goals like stars, users and citations stay goals until I earn them.

I welcome advice from researchers, students and engineers, especially in Urdu NLP, multilingual AI and trustworthy RAG. If you supervise undergraduate researchers, I would be grateful to hear from you.

📧 [email protected] · 💼 LinkedIn · 🌐 Portfolio · 🔬 ORCID