Say hi

Senior Auditor at Mercor · Founder, Sciensify

I make AI measurable.

Hi, I'm Abdullah, a data scientist in New Jersey. I audit what AI models say, build the data and GenAI systems around them, and ship software that people actually use.

See my work Let's talk
AI evaluationPythonSQLPower BIAWS Bedrock Azure OpenAILangChainRAGSnowflakeReact

About me

I've spent five-plus years across data analytics, generative AI and full-stack engineering. I've built RAG systems on insurance documents, scaled Hadoop pipelines to millions of daily records, cut AWS compute costs 40%, and shipped web apps from the database up.

Today I work on the layer most people skip: evaluation. At Mercor I review AI-generated code for 184+ developers worldwide, score what models produce against a rubric, and flag where it went wrong, because that's the step that decides whether AI is safe to put in front of a customer. I'm also finishing an MS in Data Science at NJIT.

Data tells the story. I just help it speak clearly.

  • 5+years of experience
  • 184+developers' AI code reviewed
  • 40%AWS compute cost cut
  • 6certifications earned

Evaluate

  • Response auditing
  • Rubric design
  • Agentic AI
  • Prompt testing
  • Data integrity

Analyze

  • Python
  • SQL
  • Power BI & DAX
  • Tableau
  • Snowflake
  • Spark

Build

  • AWS Bedrock
  • Azure OpenAI
  • LangChain & RAG
  • TensorFlow
  • React
  • Next.js

Things I've built

Multi-agent AI · 2025

SafeClaim AI

Insurance claims from photo intake to decision in seconds: three AI agents handle damage review, fraud screening and coverage checks, and anything unusual goes to a human adjuster. Running on AWS.

  • Python
  • LangGraph
  • AWS EKS
  • RAG
Personal command center · 2026

Homebase

One place for my projects, tasks, study plans, mail and AI agents. Claude drafts email replies and LinkedIn posts for my approval, and it runs on my PC and my phone.

  • Next.js
  • MySQL
  • Claude Code
  • MCP
Productivity · 2024

AI Task Manager

A to-do app with a thinking partner: before a task is saved, AI helps clarify the scope, write a real description and propose subtasks. You approve every step.

  • JavaScript
  • Node.js
  • Claude API
Machine learning · 2024

Recipe Traffic Predictor

Predicts which recipes will drive high traffic before they're published, so editors feature content on data instead of gut feel. 85% precision.

  • Python
  • XGBoost
  • Data analysis
Agentic AI safety · 2025

AI Code Executor

Lets an AI run the code it writes without endangering the system: extracts the code, runs it isolated with strict time limits, and returns clean, structured results.

  • Python
  • Agentic AI
  • Sandboxing
Writing & research

Sciensify

My blog on AI, evaluation and data science, written in plain language for people who build with it.

  • Blog
  • AI
  • Data
Visit sciensify.com

The journey

  1. Jan 2026 – now

    Senior Auditor · Mercor

    Auditing AI model responses against rubrics and flagging where they go wrong, including AI-generated code for 184+ developers worldwide.

  2. Aug 2025 – Jan 2026

    Senior Data Scientist, Agentic AI · Mercor

    Building and evaluating self-correcting coding agents with LLM orchestration, and red-team evaluations to make models more reliable.

  3. 2025 – 2026

    MS, Data Science · New Jersey Institute of Technology

    Coursework in deep learning, machine learning and big data.

  4. Jul 2022 – Apr 2025

    Data & AI Analyst / GenAI Engineer · Vault Insurance

    A RAG-powered Q&A system on insurance documents, Hadoop ETL scaled to millions of daily records, a 40% cut in AWS compute costs, and GenAI on AWS Bedrock and Azure OpenAI.

  5. Oct 2020 – Jun 2022

    Full Stack Developer · OCloud Solutions

    A React and MongoDB web app with JWT auth, a Power BI dashboard for public health authorities, and analytics that shaped marketing decisions.

  6. Jul 2019 – Sep 2020

    Assistant Research Analyst · FAST-NUCES

    Statistical models for predictive insight on research data.

  7. Before that

    BS, Computer Science · FAST-NUCES

    National University of Computer and Emerging Sciences.

Certifications

  • Snowflake: Data Warehousing & Analytics
  • Machine Learning with Python · IBM
  • Python for Data Science & ML Bootcamp
  • Power BI: Microsoft Data Analyst (×2)
  • Python Developer Professional
  • Microsoft AI-901 · in progress

Lately I've been writing

LinkedIn

Evaluation is the road between AI and the business

Five checks before an AI answers a client: accuracy, compliance, completeness, tone and escalation.

 Sciensify

Read the blog

Longer pieces on AI, evaluation and data science, in plain language.

Let's build something useful.

Hiring, collaborating, or just curious about AI evaluation? My inbox is open.

Email me LinkedIn GitHub