I lead product design for Sincera, an AI coaching platform I've shaped from spec to live product and continue to evolve, designing not just the screens, but how the AI coach itself thinks, speaks, and remembers.
Sincera Studio is an AI-powered personal growth coach, an always-available mentor that holds your goals, remembers your patterns over time, and meets you the way you actually need to be met. I lead the product design end-to-end, working directly with a developer on implementation, and use AI as an active design partner across the whole lifecycle, from spec to prototype to live product. It's an ongoing project that keeps evolving as real use teaches us more.
People who want to grow (in their career, creatively, or personally) often can't access what actually works: a mentor who knows their history, remembers what they've tried, and is available when they actually need them. That's a privilege reserved for a lucky few.
Sincera set out to close that gap with a personal mentor that's always available, not a generic chatbot, and not a one-size-fits-all productivity app. That created three real design problems:
A coach that doesn't adapt to how you want to be coached won't get used past the first session.
Growth is a pattern across many conversations, not one. The product had to hold that history usefully, not just log it.
Because the product is partly a conversation, much of the design work wasn't screens: it was defining how the AI itself should think, speak, and respond.
Three core flows carry someone from honest insight to real momentum.
Self-assessment. Understand where you actually are.
Goal-setting. Turn that insight into goals worth pursuing.
Roadmaps and check-ins. Keep momentum over time.
I led product design end-to-end, working directly with a developer who owned implementation. My scope covered the full product definition and experience: writing the product spec, designing the core flows (Reveal → Align → Move), iterating on UI copy, and defining how the AI coach itself should behave. AI was a working tool throughout, not just for research, but as an active design and QA partner.
I wrote the product spec conversationally, working from the core insight (people need an always-available mentor) through what it requires as a product: onboarding, goal structure, memory, coaching tone. Stress-testing edge cases in dialogue ("what if a user wants encouragement one day and bluntness the next?") surfaced the coaching-personality problem before a single screen existed.
Before committing to Figma, I prototyped the logic of each stage, Reveal (self-assessment), Align (goal-setting), and Move (roadmaps, check-ins), in conversation. That let me test the shape of a flow (what it asks, in what order, what it does with answers) before investing in high-fidelity screens. Figma was for finishing well-reasoned flows, not discovering problems mid-build.
For a product where the core interaction is a conversation, copy isn't decoration. It's the interface. I drafted and iterated copy across dozens of small moments (onboarding prompts, check-in questions, coaching responses) until the tone matched the product's intent: calm, honest, non-generic.
The most novel part: designing how the AI coach behaves: its tone options, how it references past conversations, where the line sits between "supportive" and "generic." Designing a personality, not just a layout, a discipline that barely existed a few years ago.
Once flows and copy were solid, the developer built each feature; I reviewed every implementation against the spec and flows and fed corrections back. This loop repeated continuously rather than in one big handoff: issues surfaced and got fixed in days, not sprints.
Warm & encouraging, direct & tough, or playful & light. A single default tone alienates a large share of users. Letting people curate how their coach talks to them turned a weakness (an AI can't "just know" your preferences) into an explicit, user-controlled strength.
Conversations save into categories the coach can reference, so it can surface patterns over time, the way a real mentor recalls "you've mentioned this same blocker three times." The direct answer to the continuity problem: a mentor is only valuable if it remembers.
Resolving flow logic in dialogue first meant Figma time went to finishing sound flows, not discovering problems mid-build, a process shift made possible by an AI collaborator that reasons through "what if" as fast as I can ask.
A single default tone alienates a large share of users, so the coaching personality is something people set for themselves.
A working product people can use today, and still actively being built out:
AI-era product design isn't just using AI tools faster: it's designing a new surface. A meaningful part of this work has been defining how the AI itself should behave, not just what the screens look like. That's a skill set that didn't really exist five years ago, and one I now have real, shipped experience in.
Conversational prototyping changes the economics of exploring ideas. Testing flow logic in conversation meant bad ideas died in minutes instead of after a design review, so I could explore more of the problem space per week than a Figma-first process allows.
Personalization has to be explicit when the "person" is an AI. A human mentor adapts implicitly, over time. An AI coach has to be told, which meant designing direct, low-friction ways for users to shape their own experience, like tone selection, rather than hoping the product would infer it.
