Halvren Capital, in one page.

01The firm

Halvren Capital is a one-principal, AI-augmented research desk based in Vancouver, BC, covering Canadian and U.S. operators in energy, materials, and infrastructure. Models read at scale across SEDAR+ and SEC EDGAR; the principal signs every conviction. The book is proprietary; the research is public.

The firm was founded in 2025 to consolidate research previously done in private and on the principal's own balance sheet. Halvren manages proprietary capital only; it is not a registered investment adviser, broker-dealer, or portfolio manager, and is not currently accepting outside investors.

02Performance · since 2019

03Key facts

Inception

Jan 2019

Custodian / Prime

Interactive Brokers

Mandate

Long-only equity

Geography

Canada · U.S.

Sectors

Energy · Materials · Infra

Coverage universe

31 operators

Leverage

None

Derivatives

None

Outside capital

Not accepted

Methodology

Time-weighted

Sharpe ratio

1.13

Max drawdown

≈−13% (Mar 2020)

04Founder · Amirali Karimi

Amirali Karimi is the founder and principal of Halvren Capital. Vancouver-based; Simon Fraser University economics. Holds a Canadian securities licence and is a CFA candidate. Close to a decade in financial services, corporate lending, real estate development in British Columbia, and software operating across Canada and the MENA region — including a 2023 exit to Digikala, Iran's largest e-commerce platform.

The combination of operator background and capital-allocation discipline is what informs Halvren's lens: operator-led businesses, balance sheets that survive a downturn, and the capital allocation patterns that compound across cycles. Languages: English · Farsi.

05Research published

Five full company writeups currently live, each with FY 2025 data, business analysis, and the Halvren Checklist applied. Four Canadian, one Infrastructure piece spanning the U.S. through the Mainline corridor.

06How the desk works

AI ingests every filing, transcript, and insider transaction across the 31-name universe via SEDAR+ and SEC EDGAR. Models extract structured data and flag tonal or substantive changes. The principal reads every name that earns the read, applies the Halvren Checklist (10 questions across business, people, cycle), and signs every conviction. No algorithmic trading; no model in charge of capital. Public artifacts: the weekly Digest, the per-name Research, the quarterly Letter.

07Press contact