I'm Denis.
Ex-lawyer, recovering economist, software engineer focused on data.
I write code for fun and for profit.
Born in Brazil. Turned German. Currently based in Lisbon.
Reach me at me·at·denismaciel.com.
projects
Casatoo
Finds Portuguese homes before they disappear from the marketHeraclito
Legal AI that reads Brazilian case law like a tireless associateBeluga
Turns social chatter into signals teams can act onDanton
Tracks Brazilian politics before bills become problemsPraise Penguin RIP
Made Slack praise feel a little less awkwardwork
Founding Engineer & Head of Datare:cap
ML EngineerABOUT YOU
Data ScientistDataiku
Data ScientistTandemploy
Data AnalystGeschenkidee.de
education
MSc EconomicsHumboldt University of Berlin· 1.7
Sep 2016 – Oct 2021BSc EconomicsHumboldt University of Berlin· 1.7
Oct 2012 – Sep 2015academic work
Demand Learning under Limited Inventory: a simulation-based study
Master thesis studying dynamic pricing when sellers must learn demand under limited inventory, successfully reproducing the source paper's Thompson-sampling results before showing that the edge weakens under competition
Causal Forests: How machine learning can enhance econometrics in uncovering heterogeneous causal effects
Seminar paper applying causal forests to the Oregon Health Insurance Experiment, finding that Medicaid appears to increase emergency-department use for most participants while the estimated individual effects remain too uncertain to confirm meaningful heterogeneity
Replication of Erceg & Lindé (2012): Is there a Fiscal Free Lunch in a Liquidity Trap?
Seminar paper replicating Erceg and Lindé's New Keynesian liquidity-trap model, showing that government spending can have unusually large output effects but that the fiscal free lunch result depends on trap duration, price rigidity, monetary-policy response, and debt dynamics
Backus, Kehoe and Kydland (1993) - International Business Cycles: Theory and Evidence - Have the conclusions changed?
Bachelor thesis revisiting Backus, Kehoe and Kydland's international business-cycle evidence with 23 additional years of data, concluding that the quantity anomaly still persists