Fernando G.

Clearco is a revenue-based financing platform that provides non-dilutive capital to thousands of e-commerce and SaaS businesses, using data-driven underwriting models.

• Built an LLM-based document extraction pipeline (Ollama, Phi-4, Docling, Vertex AI), reducing client document processing from days of manual review to minutes of automated extraction

• Migrated revenue forecasting for thousands of businesses from daily to weekly cadence, cutting compute time by 14% while preserving forecasting accuracy

• Implemented an uncertainty quantification framework using conformal prediction, replacing point-in-time forecasts with calibrated uncertainty bands to flag worst-performing forecasts

• Drove the analytical decision on launching a new banking-data-based financing product, providing the modelling that compared it against the existing revenue-based offering

• Diagnosed and eliminated a recurring timeout issue in the production forecasting system, removing a long-standing source of failed runs

• Designed an end-to-end automated ETL pipeline (Snowflake, Airflow, Cloud Composer) to generate datasets for forecasting research at scale

• Built a Streamlit-based annotation and metrics suite to evaluate LLM extraction quality across iterations

Zixen Trading is a high-frequency crypto trading venture co-founded with former Transfero colleagues to develop systematic trading strategies on digital assets.

• Co-founded the company and shaped its initial technical direction in systematic crypto trading

• Built the PnL tracking infrastructure that connected to multiple crypto exchange accounts and ingested live market data

• Developed the data layer used by trading strategies for backtesting and live execution

• Owned the full ML and data engineering stack as a founding technical lead during the build phase

• Designede og udviklede ETL-datapipelinen til beregning af massive finansielle datasæt, hvilket reducerede beregningstiden fra flere dage til timer (PySpark – Databricks – Azure)
• Oprettede en "metodik vedrørende kode" til at vejlede udviklingen af nye maskinlæringsprojekter, hvorfra to projekter blev struktureret;
• Udviklede maskinlæringsmodeller til at forudsige Bitcoin Trading Volume (XGBoost, 65 % nøjagtighed) og Crypto Volatility (Temporal Fusion Transformer, 95 % nøjagtighed);
• Udviklede en MLOps-pipeline til at sætte modeller i produktion: Beregning af hundredvis af funktioner, streaming af inferenser i realtid ved hjælp af FastAPI Websocket-server, modelovervågning ved hjælp af SQL-database og Slack API (Docker på ECS, AWS FeatureStore, FastAPI, SQL, Slack), med en oppetid på over 99 %;
• Oprettelse af et numerisk indeks til scoring af Time Series-data og hjælpe med at opbygge handelsstrategier (statistiske korrelationer, co-integration, ACE-score, tidsserieanalyse).

Udførte en statistisk analyse af kundeafgang for en stor brasiliansk B2C-videostreaming-app.

• Udviklede algoritmer til GPS-filtrering, korrektion og præstationsanalyse for amatørfodboldspillere, herunder lette fodboldatleter.
• Tog sig af at sikre kvaliteten af dataprodukter og forvaltningen af opgaver såsom backlog, OKR'er og målevidenskab.
• Integrerede Python-algoritmer med AWS på Serverless Framework, herunder Lambda, S3, EC2, DynamoDB og CloudWatch.
• Udførte opgaver såsom dataanalyse, implementering af Kalman Filters, Neural Networks, Fuzzy Filters, videnskabelige algoritmer, motor til statistisk indsigt, grundlæggende computervision og algoritmer til ansigtsgenkendelse.
• Brugte Python til at skabe effektive datavisualiseringer.
• Vedligeholdt Python-infrastrukturen til automatisk levering af hundredvis af daglige.

• Data Scientist from Day 1 of the company, from 0 to over 50000 users

• Algorithms for GPS filtering and correction and performance analysis fore lite football athletes and amateur players

• Client-facing customer success role regarding data services of the product

-Responsible for data products quality assurance, backlog, OKRs, and metrics

• Integrated Python algorithms with AWS on Serverless Framework (Lambda, S3 , EC2, DynamoDB, CloudWatch)

-Tasks included Data Analysis , Kalman Filters , Neural Networks , Fuzzy Filters , scientific algorithms , statistical insights engine ,basic computer vision and facial recognition algorithms

• Powerful data visualization with Python

• Maintenance of Python infrastructure for automatic delivery of hundreds of reports daily