.png&w=256&q=75)
Ali Kemal Çalak@kemalcalak
1
Events attended
1
Submissions made
Turkey
1 year of experience
About me
Ali Kemal Çalak is a software engineer based in Turkey, specializing in multi-agent AI systems, agentic AI, tool calling, and AI-powered workflow automation. He holds a degree in Software Engineering from Karadeniz Technical University and currently works as a Junior Software Engineer at Colobrix, where he contributes across the full stack — frontend, backend, and AI development. He designs and builds practical, AI-driven products and prototypes using technologies such as Python, FastAPI, Next.js, PostgreSQL, and the Agno multi-agent framework. For his graduation project, he developed DialogTR, a Turkish-focused voice AI assistant featuring multi-agent orchestration and fully on-device speech-to-text. He was also a finalist in Teknofest's Artificial Intelligence in Aviation competition — one of Turkey's largest technology and aerospace events — with the Artun Vision team, and he actively contributes to the open-source community by building and sharing developer templates and projects.
🤓 Latest Submissions
WarRoom — Autonomous Competitive Intelligence
Every founder needs to know who their competitors are, what they charge, where they are weak, and how they would react to the next move — but the answer today is either weeks of manual research or an AI summary you cannot trust because it cites nothing. WarRoom is an autonomous competitive-strategy analyst. Give it one company URL and it does the rest: it infers your market from your own site, runs live web searches to discover up to 25 real competitors — classified as direct, substitute, category incumbent, partner, or emerging — asks you to confirm the field, then scrapes each confirmed competitor's public pages (pricing, product, blog, careers, integrations, security) while respecting robots.txt. The report is evidence-linked by design: every claim carries source URLs and evidence-chunk IDs re-derived from actually scraped content, so the model cannot fabricate a citation; anything unverifiable is explicitly labeled as inference. A trust panel shows exactly how much of the report is evidence-backed — with zero fabricated citations. And it reads like a strategy analyst wrote it: an explainable threat ranking (category overlap + audience match + market proximity + momentum), full competitor dossiers with SWOT, a pricing comparison matrix, a target-company baseline, a per-competitor impact and defensive move, and a CEO briefing (move now, signals to watch, risky bets, likely competitor response). Then the hero feature: the war game — stress-test your own strategic move against an AI competitor persona grounded in that competitor's real scraped data. Under the hood: FastAPI, PostgreSQL + pgvector, Redis SSE streaming, a Next.js 16 front end, and hybrid LLM routing — high-volume extraction on a cheap local tier, heavy reasoning on the remote tier, every call logged to a live cost panel. The whole dossier exports as a branded, evidence-linked PDF.
13 Jul 2026
