Case studies
Calculation and record-keeping systems in place of spreadsheets: a relational schema instead of drifting files, permissions enforced at the API, automated money calculations and reconciliation against a reference.
4 projects in this category · R&D and engineering
Task: Remove manual work and errors from the Excel model.
Solution: A calculation engine plus automatic reconciliation against the Excel reference.
Result: Metrics: 1:1 match with the reference across 25 report rows and 5 KPIs, 4 methodology errors caught by automated checks, memorandum issued in seconds.
Fintech · Financial modelling
Context: Investment project calculations instead of Excel: KPIs over 30 years, sensitivity analysis, export to PDF and Excel. Built from scratch on FastAPI + React.
Open project ↗Task: Eliminate data loss in spreadsheets and manual recalculation of payments.
Solution: Relational schema, API-level access control, automatic payout calculation.
Result: Metrics: 5 spreadsheets replaced by one system, manual payout recalculation fully eliminated, monthly reconciliation time down ~85%.
Backend · SaaS · Education
Context: Replacing a set of five Google Sheets with a relational system. Tracking of lessons, students, rates and settlements. Built from scratch: FastAPI and PostgreSQL on the backend, React and TypeScript on the frontend, deployed with Docker.
Open project ↗Task: Replace manual customer research (searching maps and websites, checking registries, guessing whom to write to first) with a reproducible pipeline and a clear ranking.
Solution: The pipeline collects companies and contacts from open sources, adds official data on revenue, profit and company age, and computes a 0-100 score: niche, business size, digital need, detected pain. A separate confidence value shows how well the score is backed by data: a lead becomes “hot” only if the score is confirmed by data rather than assumptions. For letters there is an address stop-list and a daily sending limit.
Result: Metrics: a 0-100 score from 10 signals and a separate confidence scale for the score; one run over bars and restaurants in Moscow collected ~2,200 leads (up from ~200); 196 automated tests.
Data · Automation · B2B
Context: The system collects companies from open sources, finds contacts on their websites, pulls official data from registries and gives every prospect a score from 0 to 100. In the web panel a manager sees the best candidates, a draft of a personal letter and the deal status; the list can be exported to Excel.
Task: Bring the operational loop into a single window.
Solution: A single monitoring and control panel.
Result: Metrics: MTTR down ~39%, report preparation x8.3 faster, SLA breaches down ~31%.
Software · Operations
Context: Metrics, requests and process monitoring in one interface for the operations team.
Describe the task, the data and the timeline. We will propose a format and an outline plan. Reply within one business day, NDA on request.
Discuss a project