About
I joined Sully.ai in 2026 as a Context Engineer and later moved into a Member of Technical Staff role. Sully is a healthcare AI company building autonomous agents for hospitals. My work is on context engineering for clinical AI: the AI Scribe and agent harnesses across medical specialties.
I came to AI from Communication Design (HCI) and seven years in B2B marketing. Most of what I think about sits at the boundary between AI systems and the people meant to use them — clinicians at Sully, marketing strategists at 99Ravens, operators before that. The engineering and the non-technical side of that boundary aren't separable jobs for me; they're one job seen from two angles.
Before Sully, I was accepted into Antler's Winter '26 Canada Residency as an Entrepreneur in Residence, working on auto-evolving context engineering skills for coding agents. I had several LOIs but withdrew to join Sully.
Before Antler, I was AI Agent Systems Manager at 99Ravens AI in Toronto. I joined as a Prompt Designer (their first AI hire) and moved into the systems role about three months in. I built the multi-agent platform, the persona layer, the agent skills library, prompt engineering systems, and evaluation rubrics.
Before that, seven years in B2B marketing: demand generation at Insider, lead operations and field marketing at Tmob, and communications plus internal AI work at Jonas Software. Most of what I know about systems came from watching expertise get lost between briefs, agencies, and execution at that scale.
That marketer-to-builder transition started before the current AI agent wave. At Rootly, I was one of the early AI GTM builders, creating a Clay-centered AI Demand Generation Engineer workflow for incident-signal monitoring, lead scoring, and real-time personalized outbound. Clay later featured the Rootly system as a customer story. I also built BudBuddy AI, a playful cannabis assistant publicly described by the oz. as the world's first A.I. budtender.
The path from marketer to builder ran through automating my own workflows with LLMs. That started with small agents for marketing tasks, then open source, then production work at 99Ravens.
Open source: Agent Skills for Context Engineering (transferable patterns for production agents), AI Investigator (autonomous research pipeline), ReadWren (adaptive interviewer), Feed2Context (social posts to research reports), and JeezAI (AI career platform, paused). Recent recognition includes Garry Tan quote-posting my SkillOpt / SKILL.md analysis on X.
How I work
- Optimize for learning rate, not title.
- Build in public by default.
- Start with the dumbest thing that works.
- Favor structure over cleverness.
- Most decisions are two-way doors. Treat them that way.
Born in Turkey. Studied Communication Design (Human-Computer Interaction) at Ozyegin University on a full scholarship, earned by ranking in the top 0.1% (about 1,997th of 2 million) on Turkey's national university exam. Outside work I read papers, build, play poker and FIFA, and watch the Blue Jays. Some street photography on VSCO.
What I think about: context engineering, multi-agent orchestration, persona embodiment, and how to make models behave like specific experts instead of generic assistants.
Reach me at muratcan.koylan@outlook.com or via X.