Technical portfolio.

The problem

Meetings produced decisions and commitments that needed a reliable record. The opportunity was to turn recordings into structured minutes, summaries, and executive briefs while keeping a person in review.

What we built

A web application that transcribes meeting recordings and produces three document types — general summary, formal minutes, executive brief — with optional attendance context, exported as letterheaded PDF, Markdown, or plain text.

The discipline

The generated document enters a fast review path before filing. A format library keeps the output consistent with the organization's own writing conventions.

The result

The recurring task of writing the record becomes a review-and-approval workflow. The system continues to evolve around how executives retain, retrieve, and act on commitments.

The problem

Receipts, voice notes, and text memos in English, Russian, and Spanish needed to become structured entries at the pace of real spending.

What we built

A messaging-based service: photograph a receipt or dictate a note, and a structured transaction is posted to the user's own spreadsheet. Input batching, language detection, and extraction happen inside the habits the user already has.

The discipline

Per-user authorization connects to the user's own storage, with access tokens encrypted end to end. The spreadsheet remains the system of record in the owner's account.

The result

Expense records that keep pace with spending, in a spreadsheet the owner already trusts and controls.

The problem

Independent tutors need a reliable way to track lessons, payments, and balances without turning a personal teaching practice into administrative work.

What we built

A messaging-based ledger. A tutor records a lesson or payment by voice or text; the system extracts the details, resolves the student's name across languages and nicknames, asks when a match is ambiguous, and posts the entry after confirmation.

The discipline

Names, notes, and financial fields are encrypted at the field level. Ambiguous matches are surfaced for confirmation, so the professional remains in control of the record.

The result

A reliable balance for every student, maintained in the time it takes to send a message.

The problem

Multi-unit residential buildings already collect hourly water-use data. The opportunity was to turn that dormant record into early signals: usage patterns, anomalies, and likely leaks.

What we built

An analysis system that processes meter files in batch: statistical anomaly detection, per-unit normalization, time-of-day usage profiles, and pattern recognition for fixtures — including sustained night-hour flow, the signature of a leak. Findings are delivered as a formatted report with charts and recommendations.

The discipline

Statistics carry the findings; the narrative layer is generated, then reviewed. Thresholds and baselines are stated in the report, so a property manager can see why something was flagged.

The result

Consumption data becomes an operating document for maintenance scheduling, leak response, and tenant conversations.

The problem

The Museum of Apoxyomenos in Mali Lošinj, Croatia, needed its exhibition film brought up to the standard of the artifact it presents: a significant ancient bronze shown to an international audience.

What we built

A full digital restoration was completed: color correction, noise reduction, scene-specific optimization across aerial, underwater, and studio footage, frame-level defect repair, and upscaling to 2160p. Final masters were prepared in archival and distribution formats.

The discipline

The work was structured for a European public institution: EU e-invoicing readiness, documentation for audit and procurement requirements, secure material handling, and data-handling standards around source media.

The result

A heritage film restored to modern exhibition standard: color-corrected, noise-free, upscaled to 2160p, with archival and distribution masters ready for public presentation.

The problem

Bulk tank operations generate telemetry, fill history, weather exposure, and maintenance events across dispersed assets. The opportunity was to connect those signals into an operating model that identifies asset risk, predicts service needs, and supports regional decision-making.

What we built

A digital twin architecture connecting physical assets, IoT telemetry, historical fill records, environmental conditions, physics-informed models, machine learning prediction, KPI scoring, and tiered business dashboards — from executive overview to regional operations to deployment optimization.

The discipline

The architecture defined digital model, digital shadow, and digital twin maturity levels; specified sensor and data-quality requirements; incorporated physics-informed vaporizer icing models; and translated technical indicators into business-facing reliability, delivery, maintenance, compliance, and cost metrics.

The result

The architecture was completed as a concept, system design, technical model, and executive presentation, delivered to the project team and executive stakeholders. It demonstrates the path from raw industrial telemetry to decision-grade operating intelligence for distributed physical assets.