Day job
Leading AI adoption at Lendi
Company-wide rollout. Strategy, tooling, training, KPIs. The unglamorous work of moving an org from 'we should do something with AI' to 'this is how we ship now'.
8.5 years at Lendi - frontend developer, then lead, then R&D, now AI Manager & Builder. My mandate: redefine the role of the engineer - from feature-shipper to product builder - and prove it ships. On the side I'm building jasne.ai. In 2024 I also collaborated with MAF, a Dubai mall network, on retrieval-heavy AI.
Company-wide rollout. Strategy, tooling, training, KPIs. The unglamorous work of moving an org from 'we should do something with AI' to 'this is how we ship now'.
A vertical AI product built end-to-end across product, UX, code, infra, AI integration, and distribution. The proof that the model I sell at Lendi works in the wild.
Notes, essays, internal memos. What the engineering role becomes when AI is the default collaborator - and what hiring, performance, and team shape look like next.
Generative UI as a product surface: agents that render the interface they need, and MCP apps as the bridge between tools, data, and real workflows.
Most "AI managers" arrive at AI from outside the company. I arrived from inside - eight and a half years of shipping the product, leading an 8-person team, and watching what actually breaks. That history is the leverage.
The part I'm especially proud of: helping scale Lendi from a startup into Poland's #1 broker and the fastest-growing broker in Europe.
Vue/Nuxt developer. Frontend on B2B and B2C surfaces.
Owned the front-end stack. Led an 8-person frontend team. Nuxt 2 -> 3 from alpha.
One year off the product, on the future. AI prototypes, internal tools, the company-wide case.
Lead AI adoption company-wide. Redefine the engineering function.
A vertical AI product I'm building end-to-end - design, code, infra, distribution. The thesis I sell at Lendi, proven on my own time.
The old rule was: ideas are free, execution is the moat. AI breaks that too. Building is getting cheaper; the scarce work is knowing what should exist, for whom, and why it spreads.
jasne.ai exists to help companies turn product truth into distribution: sharper positioning, faster campaigns, stronger proof, and tighter loops between product and market.
In the next few years, the unique value of human influence - taste, trust, credibility, relationships, point of view - becomes the scarce layer. jasne.ai is built to amplify it.
AI collapses the distance between idea and shipped feature. The job stops being 'write code' and becomes 'own outcomes'. That's my main KPI at Lendi.
Business logic stops being trapped inside the product. When interfaces can be generated and rewired, the valuable layer is the workflow: context, decisions, proof, distribution, and the loop around them.
Whole departments used to form because a problem was expensive to solve. AI changes the price of that problem, often drastically. Roles will change because the old org shape stops matching the new cost.
I still prefer vendor-agnostic systems. Knowing PostgreSQL directly matters here: if Supabase or Neon stops fitting, migration is a product decision, not a company trauma.
Lead AI adoption company-wide. Redefine engineering function around the product-builder model. Ship internal tooling, agentic workflows, and the cultural shift that makes them stick.
In 2024, collaborated with MAF, a Dubai-based shopping mall network, on vector search across large-scale data and knowledge bases, retrieval workflows, and practical AI surfaces for operational context.
Vertical AI product: design, code, infra, distribution. Vendor-agnostic stack on Vercel AI SDK.
One year exploring AI as the next platform. Prototypes, internal tools, the case for company-wide adoption. The bridge between frontend lead and AI manager.
Seven years on Lendi's product. Started as Vue/Nuxt developer, grew into front-end lead, and led an 8-person frontend team. Shipped Nuxt 2/3 from alpha and owned the front-end stack across B2B and B2C surfaces.
B2B financial calculators on Vue and Angular. Where I learned how much UI sits between a financial product and a customer.
Pulled from the same thread as the PostHog page: AI-to-UI, XR, tool access, sandboxed agents, realtime context, and model-native interfaces.
Applications where model context, tools, and operational systems are exposed as a native product surface instead of hidden behind a prompt.
Interfaces that assemble around the user's intent, data, and task instead of forcing every workflow through static screens.
Gemini Live, GPT Realtime, voice loops, multimodal streams, and the product patterns that appear when latency drops low enough to feel conversational.
The jump from text and image generation into spatial assets, 3D capture, editable scenes, and product workflows built around generated reality.
Interfaces that move from screens into spatial context: planning, training, support, sales, and field workflows where the environment becomes part of the product.
Agents that can plan, execute, test, and recover inside isolated workspaces: enough freedom to do real work, enough containment to keep systems observable and safe.
Operating systems and workspaces where the model is not an app inside the environment, but the layer that routes intent, memory, files, tools, and actions.
The work matters, but so does how I show up around it. I try to be a decent person: direct, fair, reliable, and useful to the people building with me.
I train boxing. It keeps the feedback loop immediate: discipline, composure, humility, and no room for pretending.
I follow cars through engineering, design, culture, and the feel of well-built machines.
I follow new interfaces, infrastructure, AI systems, XR, and the tools that change how products get built.
I like scientific thinking: models, evidence, systems, and changing your mind when reality disagrees.