Marc Kaio Hasler Contact

I'm Kaio. I study finance in Zurich, write my thesis on LME industrial metals, and build the tools I need with AI coding agents.

Available for a full-time internship from January 2027, in physical commodity trading or sustainable finance. Until December 2026 I am on exchange at Nanyang Technological University in Singapore.

Email me LinkedIn

Since 29 June 2026 my AI coding agents have processed 29.3 billion tokens. How I work with AI

Why commodities

The most interesting course I've taken so far at UZH was on commodity trading. It made me realize how much of finance I'd been seeing in the abstract, and how much sharper everything becomes when you trace it back to physical reality.

Copper actually moving between warehouses.

Vessels rerouting around the Cape because the Suez isn't safe.

Counterparties that exist as people, not just balance sheets.

I am interested in commodity markets and in sustainable finance, and I see the two as connected.

Zurich and Singapore

Zurich

University of Zurich

I study Banking and Finance at the University of Zurich, with a minor in General Economics. The degree runs from 2024 to 2027 and my grade stands at 5.3 of 6.0. The most interesting course so far was the one on commodity trading.

Singapore

NTUNanyang Technological
University

I chose the exchange semester at NTU because Singapore is one of the hubs of that trade. I am here from August to December 2026. My courses include:

  • Enterprise Risk Management for Commodity Traders
  • Fixed Income Securities
  • Sustainable Finance
  • Cultural Intelligence at Work

How I work with AI

I work inside Claude Code for everything, so the agent has the full context. High context is the most important thing. People communicate better with it, and so do AI agents.

29,253,876,717

tokens processed by my AI coding agents in Claude Code since 29 June 2026

Counted from the session logs on my laptop, on 49 active days. Status: 6 October 2026.

Tokens per day

29,219,351,801 tokens readContext the agents read and re-read: 99.88 percent.

34,524,916 tokens writtenWritten by the models: 0.12 percent. It is the white line at the right end of the bar, drawn to scale.

For every token the models wrote, they read about 846. That is what full context looks like in numbers.

From what I say to what gets built

Diverse minds avoid shared blind spots, so I never rely on one model. Each column is one of them. Each row is a step.

  1. 1I dictate

    I say what I want in my own words. I dictate it.

  2. 2Claude writes the brief

    Whatever I say first goes to Claude, which turns the dictation into a clear brief.

  3. 3Two more readings

    The brief and my original words go to Gemini and ChatGPT. They interpret it independently and sharpen the task.

  4. 4A planning model leads

    Claude Fable plans the work, together with GPT and Gemini.

  5. 5An executing model builds

    Claude Opus does the work.

  6. 6Independent reviews

    Models from all three vendors review the result, independently of each other.

  7. 7Numbers pass a logic check

    The checker uses plain logic, not an AI model, so no figure can be invented.

Everything is built to be checkable

The aim is to keep hallucinations and guesswork out.

  • Sources are cited in APA style.
  • Every source is downloaded and kept.
  • Every number is traceable to an official database or an official company source.
  • Numbers pass a checker that uses plain logic, not an AI model.

I built ymaplify with exactly this setup. See what came out of it.

What I built

ymaplify

Live in production www.ymaplify.com
The landing page of ymaplify with the headline: Links become pins.
The public landing page of ymaplify on a computer and on a phone, captured on 7 October 2026.

The need

When I travel I spend a lot of time on social media trying to find good places, and it is hard.

What I built

A travel planning web app for exactly that need. First for myself and my friends. Now more people use it.

I built it with AI coding agents in a multi-agent workflow, with automated code review.

Why it can scale

It works like a small social network. Anyone can make a map, of good cafes or good restaurants, and share it with friends.

It is also a personal record of where something was good and what you ordered.

Visit www.ymaplify.com

The metals I watch

Bachelor thesis in progress

"Cancelled Warrants as Predictive Signals for Spread Dynamics and Volatility in LME Industrial Metals"

Four metals

AlAluminium
CuCopper
NiNickel
ZnZinc

Daily observations for each metal, from July 2016 to April 2026. The LME data is licensed, so this page shows the method and my own charts, never the raw data.

How the data pipeline works

Written in Python. It builds the price, spread and warrant panel for the thesis.

Warehouse dataDaily LME totals: closing stock, cancelled tonnage, tonnage on warrant.
PricesCash and three-month prices per metal, from a Wharton data export.
Macro seriesFrom FRED: Fed funds rate, 3-month Treasury and T-bill rates, US dollar index, VIX.
SignalCancelled tonnage as a share of stock, and its daily change.
SpreadThe best-covered price series per metal, then cash minus three-month.
GapsMissing macro values are carried forward for up to three days.
MergeOne row per metal and trading day. Warehouse data and prices must both exist, macro series are added by date.
ComputeDaily log returns and realised volatility over 5 and 20 days. The nickel trading halt of March 2022 is flagged.
Master panelWith its own checks: coverage per metal, missing values, correlations.
Chart from my thesis: change in the spread over 1 to 20 trading days, in a tight and in a normal regime, with confidence bands.
Preliminary, thesis in progress. Change in the cash to three-month spread over 1 to 20 trading days, in a tight and in a normal regime.
Chart from my thesis: response of realised volatility over 1 to 20 trading days, with confidence bands.
Preliminary, thesis in progress. Response of realised volatility to a cancelled warrant shock over 1 to 20 trading days, full sample.
Screenshot to follow

NTU course planner

Tool

A tool that plans course registration around STARS, the registration system of NTU.

Image to follow

Vistaio

Startup, 2023 to 2025

A startup I co-founded: 3D real-estate walkthroughs using Gaussian splatting. We wound it down after market validation showed no clear product-market fit.

The path so far

  1. From January 2027

    Available for a full-time internship

    Physical commodity trading or sustainable finance.

  2. Aug to Dec 2026Education
    NTUNTU Singapore

    Nanyang Technological University, Singapore

    Exchange semester.

    Enterprise Risk Management for Commodity Traders, Fixed Income Securities, Sustainable Finance, Cultural Intelligence at Work.

  3. Sep 2025 to Sep 2026Experience
    ICInvestment Club
    Zurich

    Investment Club Zurich

    Macro Analyst for US markets.

  4. 2025Education
    WU Vienna University of Economics and Business

    WU Vienna

    Summer school.

  5. 2024 to 2027Education
    University of Zurich

    University of Zurich

    BA Banking and Finance, minor in General Economics. Grade 5.3 of 6.0.

    Bachelor thesis in progress: "Cancelled Warrants as Predictive Signals for Spread Dynamics and Volatility in LME Industrial Metals".

  6. Mar to Jun 2024Experience
    UBS

    UBS

    Fraud Prevention Specialist.

  7. 2023 to 2025Experience
    AIAInitiative
    Association

    AInitiative Association

    Speaker on AI in finance.

    Talks at Forum kAIpital and the International Bankers Forum.

  8. 2023 to 2025Experience
    PPARKEM AG

    PARKEM AG

    AI-assisted workflow automation.

  9. 2023 to 2025Experience
    VVistaio

    Vistaio

    Co-founder.

  10. Jul 2022 to Apr 2023Experience
    SASwiss Armed
    Forces

    Swiss Armed Forces

    Communication Specialist.

    Built a reporting framework that cut communication delays by 60 percent.

  11. 2020 to 2021Experience
    LLavour
    YES mini enterprise

    Lavour (YES mini enterprise)

    CEO.

    UBS "Most Innovative Product" award 2021, top 8 of 178 teams nationwide.

German
Native
English
C2, TOEFL 114 of 120
French
Limited working
Portuguese
Elementary
Away from the desk I take photos. Open the galleryClose the gallery
Photo to follow
Photo to follow
Photo to follow
Photo to follow
Photo to follow
Photo to follow

What I am looking for

A full-time internship from January 2027, in physical commodity trading or sustainable finance.

If that fits what you do, write to me.

marc.hasler@proton.me LinkedIn