Lead Engineer, Graph Modeling at Striff (striff.io)

Modeled codebases as typed dependency graphs with entity/relationship types and ran deterministic graph analysis (cycle detection, boundary crossings, blast-radius traversal) alongside GNN scoring of changed edges. Built documentation-grounded retrieval that ranks a team's architecture documents against structural findings and validates each citation against the served set to keep graph-derived claims traceable to source rather than asserted.

Ontology & Knowledge Graph Engineer at Fylo Technology (fylo.io)

Designed the semantic model for a production knowledge graph over scientific literature with entity types (question, investigation, evidence, claim) and typed relationships (produces, supports, opposes, addresses), including explicit constraints for efficient traversal and graph algorithms. Implemented the ontology as an executable ShEx schema contract enforced at ingestion using a phased validator loop, versioned as system of record. Built ingestion/mapping pipelines and entity resolution that converges claims across sources into a single node using a three-tier cross-document merge policy. Evaluated RDF triple stores, property graph platforms, and structured serialization against interoperability/query/ML requirements, and deployed relational GNN inference for stance scoring and link prediction.

ML Systems Engineer at Joinable (joinable.ai)

Deployed scalable model serving on Kubernetes with continuous batching, autoscaling, and scale-to-zero, including health-checked rollouts. Tuned GPU precision and throughput to control inference cost and built a multi-environment deployment abstraction to move models between self-hosted/hosted inference without application changes (with retry/fallback).

ML Systems Engineer at SkanAI (skan.ai)

Built automated ML deployment pipelines on Kubernetes (AKS), replacing ad-hoc deployments and reducing model rollout time from weeks to days. Packaged CI/CD, monitoring, and serving as reusable Terraform and Helm modules with MLflow-gated promotion, and published a reference architecture for adoption by other teams.

Senior Cloud Infrastructure Engineer at RBC Royal Bank

Owned reliability, security, and SLOs on a ~4,000 TPS digital-banking platform. Operated hybrid Kubernetes with a Google Cloud management plane, multi-tenant isolation, least-privilege access, and automated credential rotation under OSFI-aligned compliance. Established GitOps delivery patterns (Flux, Kustomize) and operational automation adopted across multiple infrastructure teams, documenting decisions so the rationale outlived the engagement.

Data Engineer at Botko

Migrated on-prem ETL pipelines to GCP using Cloud Run transforms, Pub/Sub event coordination, and GKE compute. Added validation to ensure completeness across overlapping and inconsistent sources.

Founder & Principal Consultant at HADI Technology

Independent contracting practice delivering production ML platforms, MLOps automation, and cloud infrastructure. Led ontology/knowledge graph engineering engagements (semantic modeling, typed relationships, constraint-driven ingestion and validation, entity resolution and cross-document merge) as well as graph analytics evaluations and ML integration for downstream retrieval and predictions.

ML Infrastructure Developer at BorealisAI (RBC)

Built infrastructure between research scientists and production at RBC's AI research lab. Partnered with data scientists to productionize containerized ML workloads, pipeline orchestration, storage strategy, and deployment patterns; supported training behind Aiden, RBC Capital Markets' deep-RL trading agent. Installed and configured SLURM across NVIDIA DGX GPU clusters to distribute large-scale PyTorch/TensorFlow training across competing research workloads. Built CI/CD and model lineage with OpenShift pipelines integrating Jenkins, Artifactory, and Vault-backed short-lived credentials for commit-to-deployment traceability under bank governance. Led an initiative deploying S3-compatible object storage on PureStorage AI-Ready Infrastructure, improving internal ingestion throughput by over 100% so training jobs stopped waiting on data.