AI & machine learning engineer with 9+ years building agentic AI, LLM pipelines, and production ML across insurance, cybersecurity, finance, and energy. I also teach — 1,400+ professionals mentored into the field. I own ambiguous problems end to end and weigh reliability, latency, and cost as heavily as accuracy.
Every role below ends with something running in production — not a notebook.
Building agentic AI for insurance document intelligence. Shipped an LLM-powered extraction system that cut manual data entry by ~55%, and a multi-agent evaluation pipeline (LangGraph supervisor + LLM-as-a-Judge) verifying outputs at ~90% accuracy in production. Researched and benchmarked PII guardrail techniques to generate compliant few-shot examples from sensitive data.
Co-founded an AI-powered real estate platform. Owned architecture end to end, built LLM-driven recommendation and semantic search, and led a team of 4 engineers from concept to launch.
Sole ML engineer for a VPN serving millions. Shipped a malware domain classifier at 96% accuracy, optimized inference 100ms → 10ms, ran real-time anomaly detection across millions of DNS events, and owned the full MLOps stack — MLflow, Prefect, Docker, CI/CD, Prometheus.
Architected data ingestion and engineering systems for financial risk analytics at one of Canada's largest banks, using Hadoop, PostgreSQL, and PySpark.
Statistical modeling on vehicle telematics with Databricks + Azure, and a PyTorch document-layout model that automated text extraction from 15,000+ drilling reports, cutting processing time by 70%.
Production systems with real users, real constraints, and numbers I'll happily defend in an interview.
LangGraph supervisor orchestrating specialized validation agents with LLM-as-a-Judge scoring — layered context checks against imperfect ground truth, in production.
Automated field & header extraction from home and auto insurance documents — prompt engineering, hierarchical sub-prompting, and evaluation baked in.
Multi-source feature fusion — lexical, geo-IP, WHOIS — classifying malicious domains for a VPN serving millions. Research published on the company blog.
Real-time Isolation Forest + LOF pipelines over live DNS query streams, surfacing threat patterns and behavioral anomalies as they emerge.
Benchmarked confidentiality techniques for sensitive insurance data, then used the winner to generate anonymized few-shot examples that improved extraction accuracy.
Text + vector retrieval combining TF-IDF, LSA, and BERT embeddings over Elasticsearch and FAISS, with inverted indexes and LSH for fast lookup.
I teach AI and emerging technologies as an Associate Professor at Lambton College — in person and online — where I redesigned the curriculum to industry standards so students graduate job-ready.
Through Great Learning, I mentor professionals in programs affiliated with MIT Professional Education and UT Austin's McCombs School of Business, and I coach 1-on-1 on Topmate, where I rank in the top 1% of creators in North America.
Always happy to connect — whether it's collaborating on AI projects, sharing knowledge, speaking, mentoring, or just a good conversation about agentic systems and where this field is heading.