Featured
TenderSRB: AI SaaS for public-procurement monitoring
Founder & AI Solution ArchitectLive · open beta
A working multi-tenant platform, built with AI and running on AI agents, that monitors public tenders in Serbia and the Western Balkans and surfaces the relevant ones automatically.
Claude APIPostgreSQL 1634 ADRs7,700+ tests
Details
- Live in open beta, free during the beta.
- As of 20 October 2026: 60,500+ tenders processed, 28,700+ notifications delivered, 7 live data sources.
- LLM pipeline (Claude API) reads the tender documents on request and writes structured summaries; AI also translates tender cards.
- Tender matching by CPV code, region, budget, deadline and keywords on PostgreSQL 16; semantic matching on pgvector is the next step. Delivery via a Telegram bot, email and a web cabinet.
- CI/CD with automated AI code review; 34 ADRs and 7,700+ automated tests (20 October 2026).
- Second system: six tender bots on cloud infrastructure: monitoring of four procurement platforms, a combined digest, AI analysis of technical specifications and automated supplier requests.
- Architect and delivery lead across product, architecture, delivery and QA: I set the boundaries, assign tasks to people and AI agents, accept the result and own the assembly into a product.
Architecture: FastAPI · PostgreSQL 16 / pgvector · Docker · Ubuntu 24.04 VPS · CI/CD with AI review · LLM (Claude API) · Telegram bot · email
Read the case study →
Public code on GitHub →
agent-ops: open source kit for agentic delivery
How TenderSRB is builtMIT · public
The way TenderSRB is built, published as a small MIT licensed kit. A coordinator agent briefs worker agents and does not accept a result until it runs the checks again.
751 merged PRsClaude CodeHooks and gateMIT
Details
- A coordinator agent checks the value of a task and writes a short brief.
- A worker agent writes the code and the tests in its own branch and opens a draft pull request.
- Nothing is merged until the coordinator runs the checks again and verifies the live system.
- I own the decisions, the budget and production data. High risk changes are merged only by me.
- Each rule in the kit comes from a real incident. Eight of them are described in the repository.
- 751 merged pull requests between 2 May and 20 October 2026.
Contents: worker launcher · two hooks · three skills · four short essays · tests
See the code on GitHub →
Featured
Two-Site Data-Center Architecture and Migration
Own IT integration consultancy · flagship project
A two-site architecture for critical IT infrastructure: a metro cluster over dark fibre with synchronous storage replication.
10 sites scored132 units surveyed47 relocated30-min window
Details
- Designed a two-site architecture for critical IT infrastructure. Ten facilities scored against nine weighted criteria; the winner came in at 0.911.
- Surveyed and documented 132 units and 148 connections across six racks before a single cable was pulled; 47 units relocated, the rest kept running in place by design.
- Metro cluster across two sites over dark fibre, with synchronous storage replication and a new network core paired with the old one. RTO 15 to 30 minutes, RPO 15 minutes, inside a 30-minute downtime window. Disaster recovery plan written and tested.
Stack: VMware vSAN · Veeam B&R · Cisco Nexus (LACP) · dark-fiber metro-cluster · HA/DR · RTO/RPO
How the target site was chosen (video) →
Core-Banking Platform: A Bank Built From Scratch
Managing Director, IT, major commercial bank · 2002-2010
Built the bank's entire IT from the ground up with a team of ~90: core banking, card processing, an ATM network, 12 branches and two data centers.
Team ~9012 branches2 data centersRTO 10 min
Details
- Established a PMBOK-based PMO that cut branch-opening cost ~19% and timelines ~22%; a technology committee reduced time-to-market ~24%.
- Implemented HA clustering with a 10-minute RTO and host-to-host integration with the card processor.
Scale: Core banking (Equation) · Card processing · ATM network · 12 branches · 2 data centers · HA clustering (RTO 10 min)
City-Scale Utility-Billing Platform (4M+ Accounts)
IT Director, large municipal IT operator · 2011-2012
Ran IT for a billing platform covering ~80% of a major capital's housing stock, 4M+ customer accounts.
4M+ accountsDC relocationDR rebuild24/7 ITIL
Details
- Led a data-center relocation and a full disaster-recovery rebuild; launched a new data warehouse.
- Ran a 24/7 ITIL service desk with proactive monitoring; web-services integration with city systems; data-protection compliance certified (GDPR-equivalent).
Scope: DC relocation · DR rebuild · Data warehouse · 24/7 ITIL service desk · Zabbix monitoring · Data-protection compliance
Integrated physical-security & infrastructure delivery
Own IT integration consultancy
CCTV, access control and fire & alarm systems with seamless WiFi, often as one converged system per site, from survey to handover.
93 projects>$2.4M peak turnoverFull lifecycle
Details
- Projects integrating CCTV / video surveillance, access control (PACS), and fire & alarm systems with seamless WiFi, frequently as a single converged system per site.
- Across industrial sites, residential complexes and corporate offices, each with its own physical layout, compliance and uptime requirements.
- Owned the full delivery lifecycle: site survey and design → equipment and vendor selection → structured cabling → installation, commissioning and handover.
- Led a delivery team and managed vendors end to end; part of a portfolio of 93 delivered projects; peak annual turnover exceeding $2.4M.
Systems: CCTV / video surveillance · Access control (PACS) · Fire & alarm · Structured cabling · Seamless WiFi · Video conferencing
Server room power: two feeds, one source →
Open-source migration & cost reduction
Own IT integration consultancy
Microsoft-to-open-source transitions (RHEL, PostgreSQL, open mail stack) that removed recurring licensing costs.
30-50% lower licensing TCORHEL · PostgreSQL
Details
- Estimated 30-50% reduction in licensing TCO over a three-year horizon, depending on the estate.
- Designed data-center and video-conferencing infrastructure; ran competitive tender selection for servers and storage; virtualization on VMware / Hyper-V.