[ Director of Web Engineering & Platform Architecture ]
[ Boston, MA · 30 yrs of platform & web engineering ]
[ IBM · CyberArk · Alteryx ]

I build AI Platform Engineering infrastructure — the layer that sits between agents/models and the SRE, SecOps, FinOps, and platform teams who have to keep them running in production.

Most enterprises ship AI features by stitching together notebooks, vendor consoles, and Slack channels. That works at demo scale. It does not work when you have 8 MCP servers, 12 agent fleets, 6 RAG collections, three model providers, a $50K monthly LLM bill, a CISO asking "are we OK right now," and a CFO asking "are we on budget."

This portfolio is a deliberate platform, not a pile of side projects. Nine repos. Nine surfaces. One doctrine: every layer of the AI stack needs the same governance discipline that database, network, and identity surfaces already have.

#	Repo	Surface	Question it answers	CI
1	mcp-sentinel	Tool calls	What MCP tools are exposed and how risky?
2	rag-sentinel	Retrieval	What is in the vector store and how trustworthy?
3	agent-codex	Decisions	Under what policies are decisions allowed?
4	agent-eval-arena	Pre-prod	Should this model promotion ship?
5	agent-router	Runtime routing	Which model does this request actually hit?
6	agentobserve	Runtime telemetry	What did agents actually do?
7	shadow-ai-detector	Egress	Who is leaking what to whom?
8	ai-finops-radar	Finance	Are we on budget — and why not?
9	kinetic-flightdeck	Operator	Are we OK right now? Who do I call?

                       +-------------------------------------------------+
                       |             kinetic-flightdeck                  |
                       |    (operator surface - single pane of glass)    |
                       +-------------------------------------------------+
                                            ^
        +---------------+--------------+----+----+----------------+----------------+
        |               |              |         |                |                |
        v               v              v         v                v                v
+--------------+ +--------------+ +--------+ +--------+ +-------------+ +-----------------+
| GOVERNANCE    | | PRE-PROD    | |RUNTIME | |EGRESS  | | FINANCE     | | RUNTIME OBS     |
|               | |             | |        | |        | |             | |                 |
| mcp-sentinel  | | agent-eval- | | agent- | | shadow-| | ai-finops-  | | agentobserve    |
| rag-sentinel  | | arena       | | router | | ai-    | | radar       | | (runs, traces,  |
| agent-codex   | | (gates,     | | (live  | | detect | | (cost,      | |  cost, SLA)     |
|               | |  reg)       | | route) | | egress)| |  forecast)  | |                 |
+--------------+  +--------------+ +--------+ +--------+ +-------------+ +-----------------+
                                          |
                                          v
                       +-------------------------------------------------+
                       |          AGENTS - LLMs - MCP SERVERS            |
                       +-------------------------------------------------+

Every repo in this portfolio follows the same engineering discipline:

For platform-engineering / director-level hiring managers:

• Range across the AI stack - tool calls, retrieval, decisions, eval, routing, runtime observability, egress, finance, operator surface. Nine perspectives, one doctrine.
• Three buyer profiles - CTO/Platform (mcp-sentinel, agent-codex, agent-router, agentobserve, kinetic-flightdeck), CISO (shadow-ai-detector, mcp-sentinel), CFO (ai-finops-radar). One coherent platform thesis.
• Production-minded backend design - strict TypeScript, schema validation, full test coverage, CI matrix, dashboards as deliverables.
• Enterprise-ready domain modeling - auth posture, OAuth scopes, SOC 2 / EU AI Act mapping, PII patterns, compliance frameworks, accountability rollups, anomaly detection, forecasting with confidence intervals.
• Refusal of demo-quality engineering - none of these projects has placeholder logic. Every aggregator is testable, every override rule is justified, every composite score has weights chosen for a reason. Circuit breakers have correct half-open semantics.
• Platform-thinking doctrine - no project ships without a governance loop, an operator output, and override logic that respects single-signal escalation. The same discipline applied across nine different problem domains.

30 years of platform & web engineering · Sarajevo roots · Boston-based · ~30 hardware synthesizers and a music project running since 1999 (because the work does not stop at 5pm).

• LinkedIn
• Skills Page
• Medium
• Email

"The pessimist complains about the wind. The optimist expects it to change.
 The realist adjusts the sails."
                                                    -- William Arthur Ward