About

About

AI Systems Engineer in Montreal. Production LLM systems, agentic workflows, applied ML, and the open-source infrastructure underneath.

What I do

I build production AI systems. The kind that take 30 weeks to design, deploy quietly, and stay running for years without anyone noticing. LLM runtimes, multi-agent orchestration, retrieval pipelines, crawlers, recommendation systems, and the operational guardrails (retries, circuit breakers, validation, observability) that turn research code into infrastructure.

Where I do it

I'm the founding AI engineer at CanApply (NovaVidya Inc.) in Montreal (official title on payroll: Machine Learning Developer). I joined when the AI surface area was zero and built it from the ground up: a provider-agnostic LLM runtime (llm-client), a multi-agent conversational platform with 41 operators (Dana), AI-driven faculty and program crawlers that scaled coverage 3.5x in a quarter of the time, an immigration news intelligence pipeline (CanNews), and a Qdrant-backed academic recommendation system.

I'm also founder of Trellion, an AI-native hiring intelligence platform, and I'm building Persona, a learnable memory layer that decouples access frequency from semantic salience, the failure mode that hurts Mem0, Letta, and Zep on long-horizon tasks.

Research

Co-author on two papers: a humanoid robot localization paper at RoboCup Symposium 2025 (ViTHL: Vision Transformer-Based Hybrid Localization) and an urban climate dataset in Scientific Data (Nature Portfolio, 2026). I read deeply across LLM systems, agentic architectures, retrieval, and applied ML.

Background

Before CanApply I spent five and a half years as a quantitative software engineer at Sepanta Communications, building financial data pipelines, algorithmic trading strategies, Persian NLP sentiment models, and a Telegram-based analytics platform with 10K+ users. Honours BSc in Computer Science from Lakehead University (Thunder Bay, Canada).

What I care about

Mastery, leverage, and durable systems. The interesting part of production AI is the gap between "research works" and "production works." Closing that gap, on a deadline, with a small team, is where I do my best work.

If you want to talk about a role, a project, or anything in the AI systems space, reach out.