Hello, I'm

Meer Musabih

AI & Software Engineer

BS Software Engineering · ML Intern @ FlyRank AI · AI/ML

I build applied ML systems — models that rank, classify, and forecast — and practical agents around local LLMs. Currently doing ML internship work on real product data at FlyRank AI.

Explore my work View CV github linkedin email
flyrank / content-refresh ranker — run summary
model          RandomForestClassifier
trees          300                class_weight   balanced
split          client_grouped     leakage_checks passed
records        ~30,000 pages      features       44 -> 32

precision@20   0.70               baseline 0.40
precision@50   0.68               baseline 0.36
output         ranked review queue · priority tiers + reason codes
01

Experience

FlyRank AI — Machine Learning Intern

hands-on ML on production search data

My work centred on one question from the business side: which pages deserve a human's attention for a content refresh? I framed it as a ranking problem, engineered the feature set, trained and evaluated the model, and turned its output into something reviewers could actually work through.

Content Refresh Prioritization capstone

notebooks & code →

Can observed content and search-performance signals rank pages by priority for human refresh review?

Data
  • FlyRank internship warehouse, July 2026 release
  • ~30,000 content-page records
  • 44 original columns → 32 modeling features
  • Numeric and categorical signals combined
Method
  • Random Forest Classifier — 300 trees, class-balanced
  • Client-grouped train/test split; no client leakage
  • Feature engineering; target/leakage variables excluded
  • Ranking-oriented evaluation (Precision@K)
Results — Precision@K
modelP@20P@50
random forest0.700.68
rule-based baseline0.400.36
improvement+0.30+0.32
Strongest signals

impressions (90d) · avg. search position · days with impressions · content age · character count · word count · ctr · pageviews (90d) · scroll rate · days with sessions

Predictions ship as a ranked review queue with priority tiers and reason codes, so reviewers start with the pages most worth their time.

This is a decision-support ranking system. It does not claim that refreshing a page will automatically improve SEO — humans make the final call.

02

About Me

I'm Meer Musabih Saleem, a BS Software Engineering student at Riphah International University (class of 2027), currently doing machine learning internship work at FlyRank AI.

Most of what I know about AI came from building things rather than reading about them: a Random Forest model that ranks pages for content-refresh review, an n8n agent that scores job listings with a local LLM, a Twilio call center that listens, thinks, and talks back, and forecasting tools for multi-outlet retail. In between: MERN apps, a Flutter campus app, and a space shooter in a single C++ file.

If I had to name a lane: applied machine learning for decision support — ranking, classification, forecasting — with LLM-powered agents as the toolkit around it. The pattern underneath stays the same: frame the question, get the data honest, build the smallest thing that answers it, then measure whether it worked.

Degree
BS Software Engineering
University
Riphah International University
Graduation
2027, expected
Now
ML Intern, FlyRank AI
Focus
Applied ML · ranking · LLM agents
03

AI Agents & Automation

AI Job Finder Agent n8n · local llm

An agent workflow that takes job preferences in plain language and turns them into a short, ranked list of suitable openings — searching live public job APIs, cleaning up duplicates, and letting a locally hosted LLM do the judging.

  1. preferences in natural language
  2. job API search
  3. listing retrieval
  4. deduplication
  5. llm evaluation vs preferences
  6. ranking
  7. structured json out

one run: 275 listings retrieved → 271 after dedup → 6 returned as matches

n8n  /  ai agents  /  rest apis  /  ollama  /  qwen 2.5 1.5b  /  json

Built entirely on free and local tooling — no paid API keys. It ranks and returns; it does not apply to jobs on anyone's behalf.

04

AI / ML Projects

PersonaGuard AI recruitment intelligence

Analyzes publicly available social-media information to support recruitment screening. A logistic regression toxicity classifier over Sentence Transformer embeddings, served through FastAPI/Uvicorn — my first end-to-end NLP system, evaluated properly.

86.4% accuracy  ·  0.950 ROC-AUC

python / nlp / logistic regression / sentence transformers / fastapi / uvicorn

AI Call Center voice ai

An automated caller built on Twilio: speech-to-text brings the caller's words in, an LLM decides what to say, text-to-speech says it — with routing handled along the way. Real conversation loop, live over a phone line.

python / twilio / speech-to-text / llm / text-to-speech

code →

Inventory Forecasting predictive analytics

For businesses running multiple outlets (supermarkets and the like): mine current and historical inventory/sales data to forecast demand, flag fast movers and sell-out risk, and recommend what each outlet should and shouldn't stock.

python / pandas / numpy / scikit-learn / forecasting / feature engineering

code →

Microsoft Copilot Studio ongoing

Currently working through enterprise copilots: conversational assistants and business-process automation inside Microsoft's platform. Learning how AI gets deployed where non-engineers depend on it.

copilot studio / conversational ai / workflow automation

Still on the way to public repos: PersonaGuard AI, the Job Finder Agent, and SoleMate — meanwhile everything collects at github.com/MeerMusabih.

05

Software Engineering Projects

SoleMate

MERN shoe e-commerce — catalog, cart, checkout, the full stack.

react / node.js / express / mongodb / redux

Campus Companion

Cross-platform mobile companion app for campus life.

flutter / dart

code →

C++ Space Shooter

A playable shooter in a single C++ file. Fundamentals, game loops, collisions.

c++

06

Technical Skills

Core — Applied ML

machine learning · predictive analytics · nlp · classification · ranking models · feature engineering · model evaluation

Generative AI & Agents

generative ai · llm applications · ai agents · ollama · qwen · anthropic / claude

Python & ML Stack

python · pandas · numpy · scikit-learn · fastapi

Also Comfortable With

c++ · javascript · react · node.js · express · mongodb · flutter · git / github · rest apis

07

Education & Certifications

Riphah International University

BS Software Engineering — expected graduation 2027

Certifications and AI training

Anthropic

20 AI/Claude-focused courses

IBM SkillsBuild

Artificial Intelligence Fundamentals

Cisco

AI Fundamentals

08

Let's Connect

Recruiting for an AI/ML role, exploring a collaboration, or want to argue about ranking metrics? Best reach is email — I answer fast.

email meermusabih.work@gmail.com linkedin /in/meer-musabih github @MeerMusabih cv résumé (pdf) booking schedule a call

Or send a message

More coming soon — the FlyRank capstone write-up, its official completion badge once approved, new experiments and writing.