Virko K.

I build AI features that survive production, not just the demo. Six products I built end to end are live right now: Typeza, SiftFirst, AuditRush, PrepDoc, RoleNorth and VenomAnalytics. Three are my own, three I built for clients as the only developer on the project. Links to all of them are in the Portfolio section below. What I do: → LLM features that hold up: multi-model pipelines, structured outputs, retries, evals, spend caps → Python backends: Django, FastAPI, Celery, PostgreSQL, Redis → TypeScript and Next.js: full apps, API routes, Supabase, Stripe → AWS: ECS, S3, CloudFront, RDS, Lambda, Terraform → Real time: WebSockets, SSE, async processing Two examples: Typeza fans out 15 LLM calls per analysis and returns a finished result in under 3 minutes, with Stripe and Apple IAP billing on AWS. PrepDoc runs two models against each other over three rounds, so the report gives the reader a confidence level instead of one confident guess. Before going full time freelance: backend developer at DIZZBO (logistics SaaS) and full stack at Lingvist (AI language learning, 2M+ users). On Upwork: 28 jobs, 697 hours, 100% Job Success, Top Rated. I communicate clearly, I ship on time, and I leave code the next developer can actually read. Tell me what you are building.