Skip to content

Software engineer — Staffordshire, UK

Denys
Stepanenko

Backend & full-stack. I build systems that hold up under real conditions — and I use AI like an engineer, not a shortcut.

handleezpectus
baseStaffordshire, UK
emailstepanenko.den2006@gmail.com
statusopen to SWE roles — remote/hybrid/on-site UK
work authUK — no sponsorship
Algorithmic problems solved
3,200+
Public repositories
44
Commits in one repo
1,600+
Flagship builds
9

ships end-to-end

3 commercial client engagements delivered — Upwork, live URLs

owns correctness

guarantee boundaries hand-written, chaos-tested, tested to fail

AI-leveraged, gated

parallel agents + audit pipeline — velocity without slop

design-literate

system-design docs ship with the code — decisions, not vibes

JASONDJFMAMJJASON20252026leetcode-solutions1,616algorithm-patterns286taskmanager54videohub76log-observability24hft-simulator1,289student-portal58algo-visualiser3fullstack-portfolio89forgeai215webhook-engine134guarddog13collabcanvas150
Commit windows per repo — first to last commit, GitHub API · Nov 2026

01 / AI in development

How I work with AI

I use AI agents daily — for generation, scaffolding and exploration. What I ship is what I've read, audited and pushed through a gate. The model gives speed; the process gives correctness.

human in the loopautomated01specintent + tests first02agentAI generates diff03reviewevery line read04slop auditanti-pattern hunt05guarddogexit-code gate06tests + cisuite & build07shipmerge → deployrejected → fix → resubmit

Tool

GuardDOG

Hybrid repo scanner I built for the AI-code era: regex + AST layers for free, LLM only on changed files. Findings fail the build via exit code.

github ↗

Product

forgeai

Prompt-to-live-URL generator — ~15s to a deployed site with a database, ~$0.002 per generation. BYOK, self-hostable, MIT.

case study →

Discipline

Slop audit

My standing rule: assume AI output is hollow until proven otherwise. Trace the data, count the stubs, check that 'services' aren't one-line returns. Documented as a reusable workflow.

the gate →

“AI writes the draft. I own the diff.”

The full method → /ai

02 / Selected work

Flagship builds

Nine projects that show range: low-latency C++, .NET backends, frontend architecture, devtools — and AI used where it earns its tokens.

Full project index → /work

03 / About

The short version

Backend-focused engineer with full-stack range — .NET, ASP.NET Core, PostgreSQL and Redis on one side, React and TypeScript on the other. My interest sits where software architecture, distributed systems and software that works under real conditions meet — not in tutorials.

I'd rather build something that solves a real problem, badly at first, then fix it properly — that's the thread through everything in the index above. Outside of shipped work I've solved 3,200+ algorithmic problems across LeetCode, HackerRank and Codeforces: regular hard problems are the fastest way I know to sharpen engineering judgment.

Currently going deeper into distributed systems, advanced PostgreSQL and SaaS architecture. Freelance on Upwork — three clients shipped so far. Open to software engineer / backend / full-stack roles and freelance work.

04 / Contact

Get in touch

Working on something that needs a backend that holds up — or a pipeline that keeps AI-generated code honest? Fastest reply is email.

Authorized to work in the UK — no sponsorship required