Lorenzo Leão Dotto

Full-Stack Developer · Sirtec

September 2025 – May 2026 · Brazil

Electrical grid construction and maintenance contractor. Sole developer on Torre Ativa, a field presence and operations system for crews working across dispersed grid sites.

• Built an offline-first mobile presence app for field crews working in areas with unreliable connectivity — designed for two taps and no training, because the users were electricians on towers, not office staff.
• Delivered a live supervisor dashboard with alerts, replacing a process that ran on phone calls and memory. Supervisors moved from reviewing yesterday to acting on now.
• Produced the company's first operational dataset on crew presence and movement — information that had never been captured before, now used as the basis for planning.

Stack: Next.js, TypeScript, Supabase, PostgreSQL, Vercel. Deployed and in daily use.

FinRadar

Multi-bank statement analyser with in-browser ML

Next.js 16 · Supabase · transformers.js · Recharts

Upload bank statement CSVs from any institution and see categorised spending, with no manual tagging and no cloud AI cost.

• Designed a format-adapter architecture: heuristic detection dispatches to per-bank parsers that normalise to one schema, so supporting a new bank is one new file with no pipeline change.
• Built rules-first categorisation achieving 93% coverage on the reference dataset, with an optional local AI layer — an embedding model running entirely client-side in a Web Worker (WebGPU with WASM fallback), paired with a classifier trained offline to ~91% validation accuracy on 68k transactions.
• Implemented cross-file deduplication for overlapping statement periods, historical FX conversion at each transaction's own date, and full PT/EN internationalisation.