Open to AI/ML engineering & research roles

Faizan

AI/ML Engineer & Researcher

I ship LLM systems to production. Most recently at Zscaler: an LLM/MCP automation platform now running 100+ weekly workflows. My research keeps systems like these fast, fresh, and reliable.

steady state holding

latency under SLO · live

10×
Faster runtime
~$1M+
Annual savings
95%+
Task accuracy

From Zscaler: a production LLM/MCP automation system in a 500B+/day events environment.

Experience

Where I've worked

More on About
  1. May 2025 – Aug 2025

    Zscaler

    AI/ML Software Engineering Intern · San Jose, CA

    • Architected and productionized an LLM-powered Model Context Protocol (MCP) automation system orchestrating Hadoop→DBT migrations and pipeline deduplication across large-scale data infrastructure, at 95%+ task accuracy.
    • Shipped end-to-end with LangGraph, Redis, Postgres, and AWS: 10× faster runtime, ~70% lower latency, and ~$1M+ annual engineering cost savings across 100+ weekly workflows in a 500B+/day events environment.
    MCP LangGraph Redis Postgres AWS
  2. Sep 2024 – May 2026

    Pennsylvania State University

    Graduate Research Assistant, Data Science · Malvern, PA

    • Led end-to-end development of applied ML, NLP, and agentic AI systems over 10M+ record, multi-source datasets.
    • Built and deployed NLP pipelines (scraping, embeddings, supervised ML) over 1M+ text records at ~85–90% classification accuracy across energy, healthcare, and security domains.
    • Developed predictive and causal models, improving accuracy ~19% through feature engineering, PCA, and rigorous experimentation.
    Python NLP PyTorch Causal ML
  3. Feb 2023 – Aug 2024

    Beam AI

    Data Analytics Team Lead (GTM) · Berlin, Germany

    • Built production reporting pipelines and executive dashboards (Python, SQL, ETL), cutting reporting turnaround by 40%.
    • Engineered a recommendation system (collaborative filtering, Scikit-learn), improving precision 30% and product conversions 12%.
    • Developed predictive ROI models (Pandas, NumPy) supporting 5 enterprise PoCs (~€150K total contract value).
    Python SQL Scikit-learn ETL
  4. Jan 2021 – Dec 2022

    Daraz

    Data Analyst, Operational Excellence · Lahore, Pakistan

    • Led operations analytics across 50+ First Mile logistics stations, analyzing 1M+ customer-feedback records (SQL, Python) to surface systemic bottlenecks and drive an 18% increase in nationwide customer satisfaction.
    SQL Python Analytics

Selected research

Papers & systems

All research

Selected projects

Things I've built

All projects

STAR + FAR

STAR + FAR

Research

Continual learning for LLMs that stays fresh without forgetting

  • +4.1 pp Freshness (FFI)
  • +3.2 pp Legacy retention
  • 1.7 min Daily update cost
Python PyTorch LoRA adapters sparse routing (2-layer MLP)
Prototype

Real-time, explainable multi-agent AI for tornado disaster response

  • 0.86 Radar ROC-AUC
  • 0.99 Tweet macro-F1
Python PyTorch / TensorFlow LangGraph DistilBERT + LoRA

Recognition

Honors & press

Full list on About
  • Fox Scholar Award Penn State Fox Graduate School
  • Outstanding Student Achievement Award 2025–2026
  • Warren V. Musser Fellow
  • Campus Ambassador Penn State Great Valley

Now

What I'm focused on

Presenting SAGE at IEEE CoDIT in Bari this July, while our continual-learning work (STAR+FAR) is under review at ACM TIST. Both are pieces of my thesis: a production LLM pipeline that stays fresh, meets its SLOs, and corrects itself after deployment.

I've wrapped up my M.S. in Data Analytics at Penn State (4.0 GPA) and I'm looking for AI/ML engineering and research roles where real-time LLM systems are the job, not a side quest.

Writing

Recent notes

All writing
· 5 min read

Steady State

Six months, a model release most weeks, and the ground never stopped moving. The through-line: production AI stopped being a modeling problem and became a control-systems problem.

LLM systems control systems reliability
· 5 min read

The Frontier Is Now a Menu

GPT-5.6 shipped as three tiers. Claude comes in Fable and Sonnet. The labs unbundled 'the best model' into a price-quality menu, and your new job is per-request capital allocation.

LLM systems routing economics

Open to AI/ML engineering & research roles

Let's build reliable AI systems together.