BARAN.EXE
AI ENG // LVL 04
PICUS SECURITY
AI & Data Engineer · Founder
AI & Data Engineer building at the intersection of LLMs, data pipelines, and real products. Currently at Picus Security, working on cybersecurity analytics with AI. On the side: promptarena.app, flyingpapers, and whatever demands my attention next.
▸ Not looking — always up for a coffee
A real-time arena where you negotiate against AI opponents — and get scored by an AI judge.
Built solo from scratch. Users go head-to-head with GPT/Claude/LLaMA opponents in real-time negotiation sessions over WebSockets. A pairwise + rubric-based AI judge system scores each round. Stack: Next.js 15, React 19, FastAPI, Socket.IO, Supabase, Stripe.
A gamified relationship companion for couples — now with 50k+ users across iOS and Android.
An interactive daily ritual for couples. Partners get synchronized prompts, reflect independently, then unlock each other's answers. Built native on iOS and Android with real-time sync, streaks, progress tracking, and dynamic status systems that turn emotional check-ins into a habit.
The AI-native document collection platform for regulated workflows — visa files, apostilles, contracts.
Multi-role platform (User, Reviewer, Admin) with AI-powered questionnaire generation, canvas-based PDF editing via Fabric.js, and a structured file upload/translation workflow. Built for processing official documents across 20+ countries.
Personal, sharable pages for the things you love (and the people you love).
A delightfully simple builder for single-purpose love letters, anniversary sites, and tribute pages. Hyper-minimal UX, beautiful templates, instant publishing on a clean URL.
A voice-note and task management mobile app built specifically for ADHD brains.
React Native app with voice-note capture, structured task flows, and a UX designed around ADHD cognitive patterns — low friction capture, minimal context-switching, and progressive disclosure. Built with Expo Router and Supabase.
An autonomous content engine — from idea to published, no hands required.
A fully automated media production pipeline built for scale. Orchestrates script generation, AI voiceover, visual rendering, and multi-platform distribution through a single workflow. Designed to run quietly on infrastructure you control, not platforms that can pull the plug.
Predicting age, gender, and race of social media users from text — TÜBITAK-funded research.
Bachelor thesis evolving toward journal submission. Twitter/X scraping pipeline → LLM-based topic & sentiment labeling, with BERT/BERTweet baselines. Public Hugging Face Space for labeling and inspection. Recipient of the national TÜBITAK 2209-A undergraduate research grant.
