Machine Learning , Artificial Inteligence and Math enthusiast

Hi, I'm Ali Nawaf

Computer Science student passionate about creating innovative solutions, artificial intelligence, and building impactful software applications.

View ProjectsDownload Resume
Live AI Agent

Agentic Twin

An AI agent with access to my resume, project repos, research and coursework. Chat or talk by voice in real time, or run its RAG tools yourself with / commands.

Try the Agentic TwinHow it's built

Skills & Expertise

Focused on machine learning and AI systems. Click a skill to see the projects that use it.

All projects Resume

LLMs, RAG & Agents

Production retrieval pipelines, tool-calling agents and real-time voice agents.

RAG1LLM Agents1LangGraphLlamaIndexAI SDK1LiveKit1pgvector1Snowflake Cortex
Raised RAG recall from 50% to 94% at Trek Health

Used in

Agentic TwinDispatchSSRP Safety Oracle

AI Safety & Interpretability

Evaluating agent behavior and probing model internals.

Interpretability1Activation steeringAgent evalsCaptum1PCA1

Used in

SSRP Safety OracleAI-SWE Safety LabAgent Issue Safety Classifier

Deep Learning & Vision

Training and deploying models for medical imaging and signals.

PyTorch6TensorFlowResNet1SAM22Diffusion1VAE1Core ML
Found a new ICU mortality biomarker (HR = 1.22) at Houston Methodist

Used in

Cardiovascular Age from ECG & VCGAI Medical LabelerDiffusion Models as Markov Chains

Statistics & Math

The theory behind the models: a secondary major in mathematics.

Probability1Stochastic calculusLinear algebraSurvival Analysis1scikit-learn1NumPy1

Used in

Expectation-Maximization from ScratchDiffusion Models as Markov ChainsBayesian Variational Autoencoder

MLOps & Cloud

Shipping models and data pipelines to real users.

Docker2AWS1AzureKubernetesSLURM / HPCFastAPI2Linux
Automated an Azure pipeline for 10,000 records a month at Eaton

Used in

AI Medical LabelerMedAnnotateAgentic Twin

Languages

Python first, with systems and app languages where they fit.

Python11CSQLTypeScript4JavaSwift1

Featured Work

Selected AI safety and medical AI work

AI Safety · Agent Evaluation

Private research

SSRP Safety Oracle

A benchmark that grades autonomous coding agents on safety as well as correctness, built on top of SWE-bench style tasks.

  • Passive safety telemetry for DeepSWE agents running on mini-swe-agent: agents run unrestricted while oracles observe silently.
  • Two orthogonal grades per trial: the functional test result and an independent safety report.
PythonSWE-benchLLM AgentsDockerShell
Code available on request

AI Safety · Interpretability

Research

AI-SWE Safety Lab

Activation-steering experiments that locate the "refusal direction" inside open-weight LLMs and inject it at inference time.

  • Extracts refusal-vs-compliance directions from hidden states of TinyLlama-1.1B and Qwen2.5-Coder-3B across layers 10–25.
  • Steers generation with forward hooks and measures over-refusal on harmless coding prompts.
PythonPyTorchHugging FaceInterpretabilityPCA
Code

Medical AI · Deep Learning

Research

Cardiovascular Age from ECG & VCG

Research at Houston Methodist estimating cardiovascular age from 12-lead ECGs and 3D vectorcardiograms.

  • Trained ResNet50 models on MIMIC-IV and PTB-XL to estimate cardiovascular age and structural cardiac health.
  • Found the VCG-derived age gap to be a stronger ICU mortality biomarker than standard ECG (HR = 1.22, Cox proportional hazards).
PythonPyTorchResNetCaptumSurvival Analysis
Code
View All Projects