Hi, I'm Ali — I build production-scale AI that moves the metrics that matter.
Lead AI/ML Engineer with 10+ years architecting transformer-based recommendation, RAG ranking, personalization, and large-scale information retrieval. Currently at Procore, where I shipped the RAG + contextual-intelligence recommendation stack and an end-to-end LLM Product Insights Platform. Previously led ranking at Expedia and semantic supply-chain retrieval for Dell.
Impact in numbers
// shipped, measured, in productionWhat I'm working on
// current focusRAG + contextual-intelligence recommendations
Retrieval and ranking on transformer recommenders with an additional contextual-intelligence layer — topic modeling, taxonomy extraction, semantic modeling with LLMs — to lift recommendation quality and downstream conversion for Sales and Customer Success teams.
LLM Product Insights Platform
End-to-end pipeline turning unstructured + structured customer signals (call notes, tickets, reviews, telemetry) into queryable, source-attributed findings. Llama 3.3 70B structured extraction · Vector Search · a read-only MCP server queryable from Cursor, Claude, and Slack.
Generative retrieval & transformer ranking
HSTU, SASRec, BERT4Rec, TransAct, DIN/DIEN/SIM, BST; generative retrieval with TIGER and semantic IDs; Two-Tower deep retrieval-ranking; multi-task heads (MMoE, PLE); position-bias and counterfactual evaluation.
Serving at scale
Bridging model development with online serving — distributed training, streaming / real-time inference, low-latency online services, A/B testing, MLflow, Airflow, GPU clusters on GCP / AWS and on-prem.
Selected experience
// 10+ yearsProcore Technologies
Lead Machine Learning Engineer. RAG + contextual-intelligence recommendations; LLM Product Insights Platform; MCP server for NL-to-SQL analytics.
Expedia Group
Staff → Senior → ML Scientist. Shipped Expedia's first personalized flight recommender, Web Flights LTR, predictive cache-fill, CV stack for Vacation Rentals.
C5i · for Dell
Lead AI/ML, Sr. Manager. Two-stage SBERT bi-encoder + cross-encoder re-rank stack and an LLM-based entity-resolution pipeline for Dell.com supply-chain search.
More on the CV page — including Quotograph, UTSA Open Cloud Institute, the U.S. Patent on workload-aware multi-cloud scheduling, and selected IEEE / Springer publications.
Recent writing
// notes from the lab-
24 — System Design Capstone: An End-to-End Modern Recommender
-
23 — Closing the Loop: Feedback, Drift, and Continual Learning
-
22 — Feature Stores & ML Infrastructure for RecSys
-
21 — Serving Architectures: Batch, Real-Time, and Streaming
Retrieval, Ranking & Recommendation
A hands-on tour of the full recommender stack — features, retrieval, ranking, serving, and closing the loop, with code, math, and production notes. →