Juliu Pantia

Requirements · process & schema design · PostgreSQL · LLM agents · AI-assisted dev (Claude Code, Cursor)
Build production AI systems and internal tools end-to-end: elicit requirements, design business processes, data schemas and architecture — and deliver via advanced AI-assisted development. Selected delivered work:
Agentic analytics — an agent that builds churn & revenue reports over data; public demo for Gauss Money (github.com/umpmlv/churn-revenue-reporting-agent).
Multi-agent B2B lead pipeline — source parsing → extraction → drift-guarded LLM classification → RAG matching → proposal generation → CRM. Designed the process and data schema, built the async backend and an operator dashboard; accuracy treated as a critical attribute.
Management-accounting platform (service business, 50+ people, 5,000+ ops/month) — bank / CRM / cash / crypto / receipts integrations, append-only operations journal, P&L by segment, payroll; a custom MCP server connecting Claude to a live PostgreSQL. Result: bank reconciliation cut from 1–3 days to 5 minutes, zero payroll errors, ~15 h/week saved for the owner.
ASIC-fleet observability — telemetry ETL, Prometheus remote-write to Grafana Cloud, anomaly alerts, real-time dashboards. Crypto-analytics ETL — on-chain async-Python pipeline with scheduled aggregations and indicators.

Financial calculations · data-schema design · investor-cabinet requirements · Python · Django · PostgreSQL
Collective-investment platform for commercial real estate handling real money: pooled capital is invested into a property, split into shares, monetized, and proceeds returned to investors. ~200 investors, up to 70 objects and ~700M ₽ at peak; ~1.5B ₽ total turnover. Owned the correctness of financial calculations and the yield engine.
Designed and built a fair-yield engine: investors enter on different dates yet each receives the same effective annual yield — gains allocated by individual holding period. Removed the inequity between early and late entrants and the incentive to delay entry.
Gathered requirements and built the investor cabinet end-to-end: deposits, share allocation, sub-object lifecycle, reinvest/withdraw, monthly position reporting.
Designed and migrated an evolving share schema in PostgreSQL; background calculations (Celery + Redis), reconciliation against external payment data.

Negotiations with clients · requirements elicitation · CRM · participation in automation rollout
Ran negotiations with owners and investors, accompanied deals end-to-end, worked with CRM and reporting — hands-on requirements elicitation and stakeholder work in the field.
Took part in building and rolling out an internal automation system for investor and deal management that removed manual tracking and enabled scaling volume.
Result: attracted 170 investors over 2 years; investments grew from 33.5M to 350M ₽.

Financial modeling · VBA → SQL · centralized dashboards · bank-network scale
Maintained financial models used to compute unit metrics across the bank's branch network.
Automated manual calculations in VBA, then migrated them from Excel/VBA to SQL — direct queries to the warehouse instead of manual exports: faster and without discrepancies.
Took part in rolling out a centralized dashboard system: unified metric marts replacing scattered files, available to managers in one view.