Shakil Rafi home Shakil Rafi home
  • Home
  • Work
  • Experience
  • Teaching
  • About
  • Let’s talk
Generative AI · Applied mathematics · Data science

Turning complex data into useful signal.

I’m Shakil Rafi, Ph.D. I build rigorous, human-centered AI systems—from privacy-aware generative models and RAG workflows to interpretable genomic analytics—and teach the next generation to build them responsibly.

Explore selected work → Start a conversation
Teaching Assistant Professor · University of Arkansas
signal_field / live
AIGenerative + agentic systems
∑Mathematical rigor
DPPrivacy-aware modeling
→Research to decisions
10+ yrsbuilding with Python and R
6+ yrshands-on quantitative research
100+students taught in data science
2025inaugural course design award
Selected work

Built where theory meets consequence.

My work moves between research, software, and classrooms—with one standard throughout: the result should be reproducible, explainable, and useful to the people making decisions.

01 / GENERATIVE AI RAG

Responsible GenAI systems

RAG, local LLMs, orchestration, monitoring, and privacy-preserving synthetic data—designed around the problem, not the hype.

View the work →
02 / COMPUTATIONAL BIOLOGY DNA

Interpretable genomic ML

Exploratory classification of pathogen lineages using biologically meaningful genomic-island cassette architecture.

Read the research →
03 / OPEN SOURCE { }

Tools that make ideas tangible

From algebraic neural-network operations in the CRAN package nnR to genomic quality control experiments in Rust.

Explore the code →
How I work

Rigor before spectacle.

The best AI work is not just technically impressive. It is carefully measured, honest about uncertainty, and built for the environment where it must perform.

01

Frame the real decision

Start with the person, constraint, and consequence—not a favorite model.

02

Build an auditable path

Use reproducible pipelines, defensible baselines, and validation that matches deployment reality.

03

Protect what matters

Treat privacy, leakage, fairness, and failure modes as design inputs from day one.

04

Explain the tradeoffs

Translate the technical result into language a student, scientist, or executive can act on.

“A model earns trust when people can see what it knows, where it fails, and how it changes the decision.”
Working principle · Shakil A. Rafi
Let’s build something useful

Have a difficult data problem?

I’m always interested in thoughtful collaborations across generative AI, applied research, responsible analytics, and data-science education.

sarafi@uark.edu ↗ Connect on LinkedIn

© 2026 Shakil A. Rafi

Built for clarity, curiosity, and responsible AI.