Senior quantitative practitioner working across markets, software, and agent evaluation.
I spent 12 years in Canada's foreign-reserve program, including co-leading a $20 billion non-USD credit portfolio, then built DeFi trading software and protocol-risk systems. I now build and audit agent-led research across ML/RL, simulation, inference, and software automation.
- 12 yearsin Canada's foreign-reserve program
- $20 billionnon-USD credit portfolio co-led
- 10 campaignsdocumented as applied field reports
- 2.08x throughputpromoted to live transcription inference
Featured Autoresearch
- CASE 5 (2026-04-15) Tactics.md used autoresearch to train neural models that evaluated positions inside its game-tree search, testing architecture size, training targets, quantization, and self-play across 435 result bundles. Under the same 10 millisecond move budget, a model that evaluated each position 73 times more slowly still won because its judgments were better, while the model with the best validation loss played worst among matched candidates.
- CASE 9 (2026-08-06) FreeTranscribe.org doubled transcription throughput while sharing one RTX 3090 between live production and autoresearch. Error counts stayed unchanged across 295 broad cases and fell from 105 to 103 on a reviewed long-form set.
- CASE 10 (2026-08-08) Quantitative-trading autoresearch tested 1,807 unique policy-operation variants, then ran 93 mechanism-first research cycles across two long-running campaigns and completed 73 economic evaluations while sealed future data stayed untouched. A staged universe analysis screened 858 markets, showed that blindly expanding a 20-market policy failed, and grew the forward data regime from 20 to 42 markets with richer hash-bound evidence.
View the full collection (10 cases) ->
Selected Current Work
Public products, open-source infrastructure, and documented applied outcomes
- Agent Orchestration Process runs bounded unattended jobs across eight coding-agent CLIs. Its Python CLI isolates work, enforces deadlines and least-authority execution profiles, resumes exact provider sessions, validates declared artifacts, and retains durable evidence. PyPI
- Tactics.md used autoresearch to test 102 CPU-pipeline interventions, uncover eight correctness defects, and evaluate 435 neural-model result bundles. Under the same move budget, the model with the best validation loss played worst among matched candidates.
- FreeTranscribe.org is an AI transcription service with private transcript links, subscriptions, and a production-priority GPU queue. A sealed promotion doubled warm inference throughput on one shared RTX 3090 without increasing errors across 295 cases.
- FlySim is an Apache-licensed flight-control stack with a CPU reference environment and batched JAX reinforcement learning. Exact replay localized a simulation-to-live coverage gap, and retraining eliminated 14% control clipping without weakening the live safety limits.
- Clanker Analytics is a DuckDB-backed Python package for analyzing local Claude Code, Codex, and Gemini session logs. It powers my public project activity view, which reports human attention, agent resources, and automation separately. PyPI
Professional Foundation
At the Bank of Canada, I worked across trading, portfolio management, quantitative modelling, and policy before co-leading a $20 billion non-USD credit portfolio. At DELV, Euler, and Veda, I built Python trading and fuzzing infrastructure, worked on market strategy, and analyzed protocol risk using on-chain and internal data. Full resume