Pavel Osadchuk@xakpc
1
Events attended
1
Submissions made
10+ years of experience
About me
I'm a Principal-level .NET engineer and software architect with 15+ years of experience building and evolving production software, from early-stage products to complex enterprise systems. My core stack is C#, .NET, ASP.NET Core, SQL, Kubernetes, Azure, and AWS. I have led architecture and cross-team initiatives for systems processing hundreds of thousands of document comparisons across a content corpus of up to 70 TB. I remain hands-on across system design, implementation, production debugging, technical leadership, and architecture reviews. As a former founder, I connect engineering decisions to customer outcomes, delivery speed, reliability, and long-term maintainability. I’m interested in Staff/Principal Engineer and Software Architect roles involving distributed systems, developer platforms, cloud infrastructure, and technically complex backend products. I also advise .NET teams on adopting AI coding agents responsibly through better repository context, tooling, MCP integrations, testing, review, and architectural guardrails.
🤓 Latest Submissions
NøIdea Autonomous Options Trader
NøIdea is a .NET 10 autonomous options trader for an Alpaca paper account. The idea is simple: AI decides what looks interesting; deterministic C# decides what is allowed to happen. Grok researches the market and proposes a trade or NO_TRADE. Claude tries to falsify it, while two GPT reviewers independently check contract mechanics and market context. A fifth reviewer is plain C# doing portfolio arithmetic — it costs nothing, cannot hallucinate, and gets a vote. ◆ MODELS CANNOT TRADE - AI only sees a read-only Alpaca MCP server. Of 34 exposed tools, exactly 23 are allowed. If an account, order, or shell tool appears, the application refuses to start. Every write goes through deterministic C# using Alpaca.Markets. ◆ MODEL OUTPUT IS UNTRUSTED - Proposal values are checked against the market catalog before review; a discrepancy above 1% rejects the proposal. RiskGuard then validates the result again against a fresh account snapshot. ◆ EXITS ARE DETERMINISTIC - Open positions are checked every 60 seconds and exit decisions never involve AI. ◆ EVERYTHING IS AUDITABLE - Sessions, tool calls, reviews, votes, decisions, and orders are written to SQLite before money moves. The --audit command verifies the complete chain. Models research, argue, and vote. C# has final authority.
4 Sep 2026
