I turn manual processes
into production systems.
Fravalu finds the repeated work that costs your team the most hours, prices it, and builds the first fix. For businesses with no in-house IT team.
- 1 hour to 6-10 minutes per file
- 3-5 days to under 1 day
- Still in use in five markets
Results
Four things I built, and what changed.
I did this work as a Solutions Architect at Mesh Bio and in an MBA consulting project.
Clinic data, cleaned in minutes
An hour a file became 6-10 minutes, including human review.
Problem. More than thirty clinics sent CSV and Excel files with different column and package names, and a person cleaned each file by hand.
Built. An AI pipeline that turns messy files into clean, checked output. A human reviewer checks every file before it is submitted.
Measured. The first version got at least one attribute wrong in 70% of records. After the fix, 99% of the data went to the live database.
Onboarding new labs and clinics
Onboarding improved from 3-5 days to under 1 day.
Problem. Every new lab and clinic needed its own hardcoded Lambda, which meant copying code and testing incoming data against it.
Built. One pipeline driven by configuration, handling about 10,000 records an hour in Singapore. It is still in use across Hong Kong, the Philippines, Singapore, Malaysia and Indonesia.
A first overseas customer, and a team of three
The customer became a lighthouse account with follow-on regional projects.
Problem. The first overseas customer had a deadline and its own way of working.
Built. I recommended an API-based workflow, scoped it to the deadline and led three weeks of on-site deployment. I later built and led a team of three solutions engineers, who took over and expanded the APIs and pipelines.
Portfolio monitoring
Projected to improve monitoring time by about 45%.
Problem. An impact investor monitored its portfolio by hand.
Built. A five-stage monitoring workflow, designed with the client, with AI tools and agents built as proofs of concept and an MVP across all five stages, from data clean-up to reporting. I pitched the recommendations to the investor's leadership.
The offer
Scoped by the outcome, priced before work starts.
Fixed-fee audit
I log how your team spends its week, price each repeated task, and hand you a ranked list of what is worth automating. Most tasks are not worth automating, and I say so before you spend anything.
The list is yours to keep.
Fixed-price build
When the list names the workflow that pays back first, I build it for a price agreed before work starts, inside the systems you already run.
The code is yours to keep.
Engagements are invoiced through Fravalu, my KVK-registered practice in Rotterdam (KVK 42095902). I work on site in Rotterdam when a project calls for it.
How it works
I build the one thing that pays back first.
Map the work
I sit with the people doing the work and find out where the hours actually go.
Rank and price
Every repeated task gets a cost. Most are not worth automating, and I say so before you spend anything.
Build
I build with plain tools, inside the systems you already run, and the code is yours to keep.
Measure what changed
I check what it saved. If it came in under what I expected, you hear that from me first.
The first call is 30 minutes. You describe one process, and I tell you whether it is worth an audit.
Who it is for
Teams that re-key, clean or chase the same data every week.
For owners and operations leads who have no IT team to hand that work to.
About
I am Alvee. I spent three and a half years at Mesh Bio in Singapore and became a Solutions Architect leading a team of three solutions engineers. I hold the AWS Certified Solutions Architect – Associate certification. I am completing an MBA at Rotterdam School of Management.
- Languages
- I work in Python, and have shipped production C++ and Qt.
- AI and agents
- LangChain, LangGraph, Amazon Bedrock, Claude, Gemini, Pydantic, ChromaDB and n8n.
- Cloud and infrastructure
- AWS Lambda, SQS, SNS, Step Functions and Terraform.
- Before Mesh Bio
- Software engineer at a Singapore medtech company, owning the annotation and user-interaction layer of the imaging software for a cancer-ablation robot. Before that, a research assistant in robotics and wearable systems at SUTD, co-author of a paper in Advanced Healthcare Materials, 2023.
- Built on my own
- An electives navigator that puts the RSM schedule on one calendar and flags clashes. Open the navigator.
Questions
Before you book.
What do I get from the audit?
A ranked, costed list of the repeated work in your team that is worth automating, and which of it is not. The list is yours to keep.
Do I have to build afterwards?
No. The list is yours whether or not a build follows.
Who owns the code?
You do. I build with plain tools inside the systems you already run, and the code is yours to keep.
Which tools do you use?
Python, AWS (Lambda, SQS, SNS, Step Functions and Terraform), and AI tools such as Claude, Gemini, LangGraph and n8n, chosen to fit the systems you already run.
Do you work on site?
I am based in Rotterdam and work on site when a project calls for it.
Will a person still check the output?
In the clinic pipeline, a human reviewer checked every file before it was submitted, and a validation layer checked each submission against strict criteria.
Book a call
Tell me where the hours go.
Pick a time, or email me directly. I will take it from there.
KVK 42095902 · Rotterdam, Netherlands