Skip to content
Waliya KhanResearch & applied AIStart a conversation

Data science researcher · Causal inference & applied AI

Waliya Khan.

Better questions. More defensible answers.

I study how causal reasoning and machine learning can help distinguish what changes an outcome from what merely correlates with it.

Explore the researchStart a conversation
GitHub (opens in a new tab)Hugging Face (opens in a new tab)LinkedIn (opens in a new tab)
Reasoning, made explicit
From a defined question to inspectable evidence.
Data Science · In progress
MPhil
Current MPhil CGPA, supplied CV
4.00 / 4.00
University of the Punjab · BBIT
Gold Medal
01 / Perspective

The assumptions matter.

My current work focuses on treatment-effect questions in real-world cardiology data. Alongside this research, I build and evaluate agent systems, language-model adaptations and data infrastructure. I am interested in the assumptions behind a result as much as the result itself.

How I approach research
02 / Research & publication

From association
to a causal question.

A current research direction and a published study, with their scope and limitations kept visible.

Current thesis · In progress

Causal inference in
heart-failure care pathways.

Regression and causal approaches to real-world cardiology records, considering procedures and medications as interventions and echocardiogram outcomes as responses.

  1. 01Question
  2. 02DAG
  3. 03Estimand
  4. 04Identifiability
  5. 05Estimation
  6. 06Sensitivity

Current phase: mapping associations and selecting intervention contrasts. This is ongoing research, not a clinical recommendation.

Publication · iCoMET 2026

Effect of Wearable-Based Automated Interventions on Stress Reduction using Machine Learning

W. Khan, W. Munir, F. Bukhari & T. Butt

5th International Conference on Computing, Mathematics and Engineering Technologies · Sukkur, Pakistan · May 2026

The study combines a real pre-intervention wearable dataset with literature-based simulated post-intervention outcomes. Findings are bounded by those simulation assumptions; clinical use requires prospective validation.

Publication DOI (opens in a new tab)

10.1109/iCoMET69771.2026.11591768

03 / Selected projects

Ideas made
inspectable.

Open-source artifacts and applied projects spanning agent reasoning, language models and data systems.

MPhil thesis · In progress

Causal reasoning in heart-failure care

Comparing regression and causal inference using observational cardiology records with echocardiogram outcomes.

Research

  • Treat procedures and medications as candidate interventions rather than simply predictive features.
  • Map associations to decide which interventions to contrast.
  • Follow a causal roadmap: question, DAG, estimand, identifiability, estimation and sensitivity analysis.

Research in progress, co-supervised by the Founder and CEO of Quantum Healthcare (US). No clinical treatment recommendation or completed thesis result is claimed.

Tools & methods

  • Causal DAGs
  • Estimands
  • Regression
  • Sensitivity analysis
Language models · Open-source project

Agentic FinTech LLM

An instruction-tuned model built from a structured survey of reinforcement-learning security in FinTech.

Model & dataset development

  • Construct a 3,280-example instruction dataset from a 60-paper survey.
  • Fine-tune Llama 3.2 1B with LoRA.
  • Release the adapter, merged model and GGUF build for reproducible use.

Tools & methods

  • Llama 3.2
  • LoRA
  • TRL
  • PEFT
  • Ollama
Repository (opens in a new tab)
Agent systems · Interactive project

Self-reflective fact-checking agent

A drafting and auditing pipeline that exposes the steps behind a corrected answer.

Agent pipeline development

  • A drafter produces an answer and an extractor identifies its claims.
  • An auditor checks those claims against web search.
  • An editor revises the answer and retains an audit trail.

An experimental verification workflow, not a claim that generated answers are error-free.

Tools & methods

  • LangGraph
  • Groq
  • Gradio
  • Web search
Explore the demo (opens in a new tab)
Engineering · Selected projects

Data systems & applied learning

Projects exploring message pipelines and early forecasting from partial observations.

Project development

  • Build a real-time chat platform with buffered, batched storage, presence tracking and media handling.
  • Stress-test the messaging pipeline with a load generator.
  • Forecast student scores from partial assessments, with the supplied project reporting average error of about five points out of 100.

The forecasting result belongs to the supplied project evaluation, not a guarantee for a new cohort.

Tools & methods

  • Kafka
  • MongoDB
  • Redis
  • MinIO
  • scikit-learn
Chat platform repository (opens in a new tab)
04 / Approach

Evidence before certainty.

  1. 01

    Define the question

    Start with the intervention and outcome of interest, not only an available dataset.

  2. 02

    Expose assumptions

    Make the causal structure and identifiability requirements explicit.

  3. 03

    Evaluate carefully

    Separate model performance from evidence about a real-world intervention.

  4. 04

    Make work inspectable

    Keep limitations, evaluation context and reproducible artifacts alongside results.

05 / Experience

Research, teaching
and practice.

Sep 2026 — present

Teaching Assistant · Big Data Analytics

Department of Data Science, FCIT

Design and conduct labs, prepare and mark quizzes and assignments, and help update course material under the course supervisor.

Mar 2023 — Jan 2025

Teaching Assistant

IBIT, University of the Punjab

Programming Fundamentals, Object Oriented Programming and Introduction to Databases labs, mentoring and assessment.

Jul 2024 — present

CEO & Co-founder

DonDev

Applied work connecting client discovery, process design and tested automation delivery. This complements, rather than substitutes for, the academic research record.

06 / Capabilities

Methods meet implementation.

Causal inference & statistics

  • Causal DAGs · estimands · regression
  • Sensitivity analysis · exploratory data analysis
  • Observational-data questions

Machine learning & agents

  • Python · scikit-learn · LoRA · TRL · PEFT
  • LangGraph · RAG · agentic AI
  • Gradio · Hugging Face · FastAPI

Data infrastructure

  • Kafka · Redis · MongoDB · MinIO
  • MySQL · SQLite · Cassandra · Neo4j
  • Vector databases · REST APIs

Education & distinctions

Sep 2025 — present

MPhil in Data Science

FCIT, University of the Punjab

Department of Data Science. Current CGPA 4.00/4.00 in the supplied CV; thesis in progress.

Dec 2021 — Jun 2025

Bachelor of Business & Information Technology

IBIT, University of the Punjab

IT specialization. CGPA 3.99/4.00. University Gold Medal, 135th Convocation; Honhaar Scholarship and merit-based laptop award. Third place in the IBIT coding competition.

07 / Contact

A conversation grounded in the work.

For research exchanges, questions about the projects or academic collaboration, please get in touch by email.

A direct conversation

waliyashahkhan@gmail.comWrite an email

Opens your email app. No form or automatic submission.