Reconstruction: to justify SPIE’s market value, margins would need to rise from 6.7% to about 8.1%.
VerifiedEvery input traced to SPIE’s annual reports, then the valuation rebuilt independently and matched to about 1e-11.
Limit. Built after the fact, on analyst assumptions. Not a price target. Terminal value is 75% of the total.
SPIE buyout (hypothetical)
Hypothetical: buying SPIE at a 20% takeover premium, with a 12× EBITDA exit, would lose sponsor money: 0.80× back, or −4.3% a year. Without the premium it breaks even.
VerifiedRebuilt independently from its written contract and matched to floating-point precision (largest gap 2e-12).
Limit. A hypothetical scenario on analyst assumptions, reference date 31 Dec 2025. Not a transaction, not investment advice.
Independent AI audit
A second AI checked a first AI’s research line by line: 91 claims, 68 supported, 15 partly, 1 contradicted. It found the hypothetical buyout’s loss was essentially the 20% premium, which the original headline did not say.
VerifiedEvery model rebuilt independently, workbooks recalculated, sources re-fetched.
Limit. I commissioned and directed this audit. 129 supporting documents were scanned, not read line by line.
EU macro data pilot
A free data aggregator was missing most recent figures, so the official sources mattered: of 28 comparison points, 6 matched, 18 were missing and 4 differed.
VerifiedAll comparisons reproduced from the raw downloads; official values re-fetched a day later, with no revisions. Seven series came from Eurostat, DG ECFIN and the ECB.
Limit. A small sample of series. Not a reliability verdict on any provider.
Bachelor’s thesis
Synthetic control: measured a 5–6.1% decline in Russia’s GDP per capita from Western sanctions after 2014. Graded 5.75/6.
Direct. I ran multi-agent research with separate producer and reviewer agents, under written contracts.
Verify. I challenged the first portfolio as too shallow and commissioned an independent audit by a different AI. It found the buyout headline hid its main driver.
Disclose. Every project states what was checked and what wasn’t.
Study tool for the CFA® Program
Turns the 10 CFA® Program curriculum volumes into layered study notes with AI.
I research EU transport economics at the International Road Transport Union (IRU) in Geneva, building econometric models that forecast freight demand across the 27 EU member states. My bachelor’s thesis used the synthetic control method to estimate the GDP impact of Western sanctions on Russia. I combine this quantitative foundation with hands-on work in AI tools and multi-agent systems.
Experience
Junior Research Econometrician
International Road Transport Union (IRU) · Geneva, Switzerland
– present
Build econometric models that forecast freight transport demand across all 27 EU member states.
Deliver short-term and long-term demand projections used in IRU policy briefs.
Process and analyse large EU transport datasets in Python and R.
Econometric Intern
International Road Transport Union (IRU) · Geneva, Switzerland
–
Analysed EU transport industry data and built models projecting short-term and long-term transport demand.
Carried out market analysis to support business expansion.
Extracted, cleaned and analysed transport data.
Assistant to WTO Counsellor
Permanent Mission of Uzbekistan to the UN Office · Geneva, Switzerland
–
Wrote reports for the UN Office and international organisations.
Liaised with embassies and supported colleagues in WTO policy work.
Produced daily reports on socio-economic conditions across world regions.
Selected work
SPIE buyout (hypothetical)
Verified
Hypothetical: buying SPIE at a 20% takeover premium, with a 12× EBITDA exit, would lose sponsor money (0.80× back, or −4.3% a year); without the premium it breaks even.
Checked. Rebuilt independently from its written contract and matched to floating-point precision (largest gap 2e-12). The workbook was recalculated in a separate engine.
Limit. A hypothetical scenario on analyst assumptions (exit multiple, financing terms, tax), reference date 31 Dec 2025. Not a transaction, not investment advice.
SPIE cash and valuation
Verified
Reconstruction: to justify SPIE’s market value, margins would need to rise from 6.7% to about 8.1%; its FY2025 cash after leases and investment was about EUR 554m.
Checked. Every input traced to SPIE’s annual reports (page-cited), then the valuation rebuilt independently from the stated assumptions and matched to about 1e-11.
Limit. Built after the fact from reports published in 2026, on analyst assumptions. Terminal value is 75% of the total. Not a price target.
Independent AI audit
Verified
A second AI checked a first AI’s research line by line and found that the hypothetical buyout’s loss was essentially the 20% premium, which the original headline did not say.
Checked. 91 claims: 68 supported, 15 partly, 1 contradicted, the rest unverifiable, untested or historical. Every model rebuilt independently, workbooks recalculated, sources re-fetched.
Limit. I commissioned and directed the audit. 129 supporting documents were scanned, not read line by line (stated in the report).
EU macro data pilot
Verified
A free data aggregator was missing most recent figures: of 28 comparison points, 6 matched, 18 were missing and 4 differed, so the official sources mattered.
Checked. All comparisons reproduced from the raw downloads; official values re-fetched a day later, with no revisions. Seven official series came from Eurostat, DG ECFIN and the ECB.
Limit. A small sample of series. Not a reliability verdict on any provider.
Bachelor’s thesis
Graded 5.75/6
Synthetic control: measured a 5–6.1% decline in Russia’s GDP per capita from Western sanctions after 2014.
Study tool for the CFA® Program
In progress
Turns the 10 CFA® Program curriculum volumes into layered study notes with AI.
Not independently checked
Education
Level I candidate in the CFA® Program (exam February 2027)
Bachelor of Arts in Economics
University of St. Gallen (HSG) · St. Gallen, Switzerland
2020 – 2024
GPA 4.9/6.
Bachelor’s thesis, 5.75/6: Assessing the Economic Impact of Sanctions on Russia. Title page (PDF)
International Baccalaureate
College Champittet · Pully, Switzerland
2015 – 2020
Grade 34/45.
Skills and languages
Technical
Python, R, JavaScript/Node.js, Claude API, LLM orchestration, multi-agent systems, data pipelines
Uzbek (native), Russian (native), English (C1), French (B1–B2)
How I work with AI
Direct
I ran multi-agent research with separate producer and reviewer agents, under written contracts.
Verify
I challenged the first portfolio as too shallow and commissioned an independent audit by a different AI. It found the buyout headline hid its main driver.
Disclose
Every project states what was checked and what wasn’t.