Ismail Anajar. I specialize in React, Next.js, and TypeScript. I work across the backend when the product needs it, and I build reliable systems around foundation models.
A multi-role service marketplace connecting clients with local service providers. Clients discover and book services, providers manage availability and fulfillment, and administrators operate the platform through a dedicated back office. I built the system end to end: the data model, the API, and the three product surfaces they feed.
A custom page-builder platform composed of an admin dashboard, a typed Hono.js API, and a frontend template consumer. Non-technical editors create and arrange pages visually in the dashboard, the API serves the persisted layout as structured data, and the template consumer renders it on the public side. I built the system end to end, from the structured data model to the editable authoring surface to the templates that publish.
I am a frontend engineer with 7+ years of experience building production web applications, specializing in React, Next.js, and TypeScript. My strongest work is frontend architecture and production frontend engineering, and I work across the full stack when the product needs it: Node.js services, PostgreSQL, REST APIs, authentication, and system integrations.
I also build AI-powered systems, focusing on the engineering around foundation models: context engineering, RAG and retrieval, tool calling, agents, evaluation, and production reliability.
I originally studied physics before moving into software engineering. The habit it left me with is the one I still use: understand the constraints of the system first, then engineer within them.
Bachelor of Physics Science
Ibn Zohr University, Agadir, Morocco
2014 – 2017
Maintainable React and Next.js applications: clear component boundaries, deliberate state ownership, and data flows a team can reason about. Architecture that keeps a product shippable as it grows, not just buildable in week one.
Performance, accessibility, SSR, responsive UX, loading and error states, and the edge cases real users find first. Frontend work that survives contact with production.
Frontend is the specialty and the stack is the range. I build and maintain REST APIs, backend services, authentication, and PostgreSQL-backed business logic when the product needs a feature closed end to end.
Connecting applications to the systems around them: third-party APIs, authentication services, payments, maps, notifications, and business platforms, isolated behind contracts instead of woven through the code.
Dependable systems around foundation models: context engineering, RAG and retrieval, tool design and calling, agent workflows, evaluation, and production reliability.
The core craft: React, server rendering, and typed client engineering at production scale.
Enough backend to close the loop on a feature, API-first.
Relational data modeled and accessed through typed layers.
Engineering around foundation models: context, retrieval, tools, agents, evaluation.
The concerns that separate a demo from a production system.
Enough infrastructure to ship and operate what I build.
I build systems around foundation models, focusing on the engineering required to turn model capabilities into reliable product features. The models are treated as given services; the work is everything around them: context, retrieval, tooling, evaluation, and operations.
Designing exactly what a model sees before it answers: the structure, selection, and ordering of context that decides whether output is grounded in evidence or produced from guesswork.
Retrieval systems that surface the right documents for the question: indexing strategy, retrieval tuned to the task, and results that feed the model as verifiable source material.
Tools as typed interfaces a model can call, and agent workflows that act, verify, and return control to a human at the moments that matter.
Evaluation before shipping: test sets, grading, and regression checks that treat model behavior as code that can change unexpectedly and must be measured.
Latency, cost, error handling, and observability treated like any other production system, because a model API is an external dependency that can fail, stall, or drift.
Before writing components I map the data flow, state ownership, and failure paths. Architecture follows from the system it has to live inside, not from the component library.
Loading, empty, error, and edge states are part of the feature. A screen that only works on the happy path is not finished.
Frontend, API, and data layers each own their responsibilities, with typed contracts at the edges. When a boundary is clear, change stays local.
A rewrite freezes the roadmap. I move one layer at a time and keep the product shippable through every transition.
The backend, not re-computation in the client, is the source of truth for business rules. When one service owns a rule, a whole class of drift bugs disappears.
Measured and designed in from the start, not audited at the end. They are properties of the implementation, not add-ons.
AI systems are measured, evaluated, observed, and improved like any other production system. An unmeasured model call is an untested dependency.
I'm open to frontend engineering roles, product-focused development work, and challenging systems involving modern web and AI engineering.