ClinicFlow
Clinic operations platform that brings doctor availability, appointment scheduling and patient records under one role-aware system.
Open the case file
- Deployed
- SaaS
- Next.js, TypeScript, Supabase, PostgreSQL
Senior Full-Stack Developer Agentic AI Engineer
Den JansenFlores
I build systems for the hours when assumptions fail: operational SaaS, full-stack platforms and agentic AI workflows engineered to keep their answers straight under pressure.
Full-stack engineering for systems that cannot afford to lose control.
I am a senior full-stack developer and agentic AI engineer based in Muntinlupa City, Metro Manila. I build operational SaaS, secure web platforms and agentic AI workflows for teams here and abroad. The work covers healthcare operations,courier logistics andAI-assisted analysis. I own the schema, access rules, server logic and interface. A weak seam is enough to compromise the whole system.
Requirements rarely arrive complete. I trace the job from one end to the other: who touches the record, which decision follows, and where the truth lives today. Often it is split between a wall calendar, a group chat and one person's memory. The point where that chain breaks is where the product begins.
I assume every shortcut will eventually be found, usually at the worst possible hour. Access rules stay close to the data. Every state has a name. Recovery paths are designed before the demo because a system reveals its quality when the ordinary path is gone.
Working with me is straightforward. We agree on the boundaries before the build starts. Updates come directly from me. If a feature has no business value, I say so before it reaches the invoice.
The visible half is easy to recognize: a clear interface, a direct explanation and a client who knows what the system will do.
The other half stays out of sight. I read the logs, test the case nobody raised and put the system under the conditions it will meet outside the demo. When that work is done properly, the silence is the evidence.
Projects
Seven case studies cover healthcare, logistics, workforce operations, commerce and AI-assisted analysis. Each file names the business problem, the architecture, the work I owned and the current state of the deployment and source. The evidence is there to inspect.
Smiley
Independent dental practices often fall back to shared spreadsheets because practice software is priced for larger groups. A platform that serves several clinics must keep every patient's data inside the correct tenant without relying on developer memory.
Smiley serves several clinics from one application. The data layer enforces tenant scope, each practice has a branded subdomain, and records, schedules and reminders inherit the clinic boundary by default.
Open the case file
Biyahero Express
Logistics operations platform that follows a shipment through booking, dispatch, proof of delivery, COD remittance and invoicing.
One shipment record carries the full lifecycle
Booking, dispatch, delivery and remittance are stages on one record, so nothing has to be reconciled across tables at the end of a run.
Open the case file
ShiftDesk
Workforce platform combining GPS-verified clock-ins, threaded support tickets and role-specific dashboards for employees and managers.
Data Analysis Platform
AI-assisted data platform that normalises business uploads, constrains model output and returns a dashboard the application can trust.
Further case files
Full-stack and agentic AI capabilities, proven in the work
A stack list says what someone has touched. It does not say what they can carry. Every capability below points to a shipped project or to this portfolio. If the evidence is not here, the claim is marked that way.
What I can be hired to own
Full-stack web application development
Complete web systems, from PostgreSQL schemas and access rules to server logic and the React or Next.js interface.
Review this serviceOperational SaaS development
Scheduling, dispatch, records and workforce platforms with multi-tenant architecture and role-based access control.
Review this serviceAgentic AI engineering
Applications where models call typed tools, return schema-validated output and carry state through guarded, multi-step workflows.
Review this serviceData platform and workflow automation
Ingestion, normalisation and validation pipelines that feed dependable analysis and reporting surfaces.
Review this service
Product and systems engineering
I turn the real operation into rules the data can enforce.
Domain modelling
Appointment lifecycles, shipment states and ticket threads mapped from the operation itself.
Multi-tenant architecture
Per-practice isolation with subdomain routing.
Role-based access control
Role-separated portals where every user can reach only the records their role allows.
Operational workflow design
Scheduling and availability, dispatch and COD reconciliation, GPS-verified clock-in.
Backend and data
The rules stay where a skipped screen cannot bypass them.
PostgreSQL
Relational models behind scheduling, dispatch and ticket lifecycles.
Supabase auth and row-level security
Sessions, tenant boundaries and row-level policies protecting deployed systems.
Next.js server routes
The server half of five deployed SaaS builds.
Node.js and Express
The API layer behind the student assessment platform.
MongoDB
Document storage where the assessment shape changes between cohorts.
Data ingestion pipelines
Uploads are normalised and validated before the analysis stage sees them.
Frontend and interface
Interfaces built for live data, impatient users and the screen already in their hand.
React
Component systems across five deployed applications.
TypeScript
Typed contracts across the interface, data layer and API boundary.
Next.js
Six deployed applications, server and client boundaries included.
Tailwind CSS
Token-driven interface systems, including the one you are reading.
Accessible interface patterns
Keyboard navigation, native disclosures, semantic landmarks and tested contrast on this site.
AI-assisted workflows
Models do useful work inside clear permissions, schemas and failure boundaries.
OpenAI API
The analysis stage of the ingestion pipeline.
Structured model output
Schema-constrained responses the interface can render without interpretation.
Prompt orchestration
Prepared context, explicit output shapes and iteration against observed failures.
Claude API and tool use
Used in client and internal work. The evidence sits outside this portfolio, so the capability remains unverified here.
Deployment and operations
A build counts when people can reach it and the evidence still holds.
Vercel deployment
Six live deployments, each linked and status-checked on its project page.
Public repositories
Six builds have public source. The private repository is labelled without ambiguity.
Astro static delivery
This portfolio ships as static HTML with no client-side framework and no hydration.
Deployment verification
Every deployment link is opened, checked and dated before its status is published.
How I build software that survives pressure
I map the real operation, keep authority close to the data and name the failure states before launch. AI remains one controlled stage inside a larger system. Each rule below points to the work that tested it.
Map the operation before choosing the stack
The workflow writes the schema.
A clinic appointment and a courier shipment obey different rules. I trace the real process first, then choose the tools. It costs more attention in week one and prevents a careless rebuild in month six.
One appointment row with explicit status transitions in ClinicFlow. A shipment lifecycle with proof of delivery and cash reconciliation in Biyahero Express.
Own the whole chain
Responsibility runs from schema to interface.
I write the data model, server logic, access rules and interface. I also answer for how they behave together. When a system fails at two in the morning, the handoff chart is irrelevant. Ownership is not.
Every project here was built end to end. Each project page lists exactly which parts, so the claim can be checked instead of believed.
Keep authority close to the data
A missed interface check should never become a disclosure.
The database enforces row-level policies and tenant boundaries. A conditional in a component may improve the interface, but it does not decide who can read a record. That decision belongs at the boundary an attacker cannot skip.
Row-level security across role-separated portals in ClinicFlow. Per-practice isolation with subdomain routing in Smiley.
Put the model inside a controlled system
Prepared context in. Validated output out. Cost known.
The model call is one stage. The surrounding code decides what context it receives, which tools it may use, what shape it must return and how the system recovers when it fails. Intelligence without control is still a liability.
The ingestion and analysis pipeline in the SaaS Data Analysis Platform: normalise, prompt, validate the shape, then render.
Design the recovery before launch
Every state needs a name, a transition and a way back.
Operational software fails at the edges: a double booking, a delivery completed twice, cash that never reconciles. I design those paths before the demo. The edge cases are where a system reveals what it was built to survive.
Booked, confirmed, completed and cancelled as states on one appointment row. Ticket lifecycles with threaded history in ShiftDesk.
Reputation is useful. Evidence is better.
Four years of project-based engineering are documented through complete case studies, public repositories and dated deployment checks. Employer timelines, client counts and performance figures stay off the page unless published evidence can carry them.
- Years building web systems
- 4+
- Builds documented here
- 7
- Live deployments, linked
- 6
- Public source repositories
- 6
My experience comes from four years of independent and client work on complete systems. The record is project-based, and every case study states which parts were mine.
Deployment links were opened and checked on . One build, Career Path, is no longer hosted, so its page carries no live link. The record should be accurate even when the answer is less flattering.
Direct answers for recruiters and clients
Location, availability, role fit, scope and the line between full-stack and agentic AI work. The answers are written plainly and remain readable without JavaScript.
Who is Jansen Flores?
Jansen Flores is Den Jansen Flores, a senior full-stack developer and agentic AI engineer based in Muntinlupa City, Metro Manila. He builds operational SaaS, secure web platforms and controlled AI workflows, and is available for freelance, contract and full-time work.
What does Den Jansen Flores do?
Den Jansen Flores is a senior full-stack developer and agentic AI engineer in Metro Manila, Philippines. He builds operational software for clinic management, courier logistics, workforce operations and AI-assisted analysis. He owns the chain from schema to interface because weak handoffs become production failures.
What is his primary stack?
His core stack is React, Next.js and TypeScript on the front end, with Supabase, PostgreSQL and Node.js behind it. He uses Tailwind CSS for interface systems. His AI work uses OpenAI and Claude APIs with tool calling, retrieval and schema-constrained output rather than a chat box added at the end.
What does he mean by agentic AI engineering?
Agentic AI engineering means a model does accountable work inside an application. It calls typed tools, retrieves grounded context, returns schema-validated output and carries state across several steps. The serious work is the control around the model: permissions, failure handling, observability and cost per run.
What roles is he open to?
He is open to senior full-stack, product engineering and AI engineering roles. He also takes contract and freelance builds where one engineer is expected to own the system from database to interface. Full-time employment and fixed-scope work are both in range.
Is he available for work?
Yes. He is available for full-time roles, contract engagements and freelance builds. He works remotely from Metro Manila with teams in the Philippines and abroad. Email floresjansen28@gmail.com. Replies usually arrive within one business day on UTC+8.
What industries has he built software for?
His documented work covers healthcare operations, dental practice management, courier dispatch, workforce systems, e-commerce, education assessment and AI-assisted business intelligence. Each sector links to a case study with the problem, architecture, ownership, source status and deployment status stated plainly.
What does working with him look like?
The rules are set before the build begins. He traces the real workflow before choosing the framework, keeps access rules close to the data and names the failure paths before launch. Updates come directly from him. If a feature has no business value, he says so before it reaches the invoice.
How can someone review the source or the live builds?
Every project page links to its deployment and, when public, its source repository. Private source is labelled private. Offline deployments are labelled offline. Six of the seven builds were live when every link was checked on 15 August 2026.
Where does his portfolio live online?
The portfolio is published at denforge.it.com, under the DenForge name. That is the canonical address for every case study, service page and machine-readable feed. The earlier address, jansen-dev.vercel.app, now redirects there permanently and should no longer be cited.
Where is he based, and does he work remotely?
He is based in Muntinlupa City, Metro Manila, Philippines, on UTC+8. He works remotely, covers Asia-Pacific business hours and can schedule early or late calls for European and North American teams when the work requires it.
Bring me the system that fails after hours.
Tell me what breaks, who carries the cost and when you need it standing. If the problem is outside my range, I will say so before either of us wastes time.
- Usually within one business day (UTC+8)
- Muntinlupa City, Metro Manila, remote worldwide