Software architect / Co-founder & CTO, Impossible Code / Medellín
Software architecture for the AI era.
I’m Jacobo Isaza — software architect, technology consultant and co-founder & CTO of Impossible Code. For more than fifteen years I’ve built full-stack software and the infrastructure underneath it for international clients in very different industries — with clinical laboratories in US health systems as my core domain.
01What I do
Three ways I help teams ship.
Senior engineering leadership for companies building software that matters — from an AI product to the infrastructure it runs on.
01
AI-powered product development
From a promising model to a product people actually use.
Most AI initiatives stall somewhere between the notebook and production. I lead the engineering that closes that gap: scoping the product, designing the architecture, building the full-stack application and standing up the infrastructure that runs models reliably — the same way we ship for Impossible Code’s clients.
More about this →02
Software architecture & infrastructure
Foundations that keep shipping long after launch.
Architecture that fits the team, the business and the domain, not the trend. I design and build the foundations of a product — domain model, APIs, data, cloud infrastructure and delivery pipelines — so it can grow for years without a rewrite.
More about this →03
Technology consulting & fractional CTO
Senior technical leadership, without the full-time hire.
For companies that need a CTO’s judgment before they need a CTO. I help decide what to build, what to buy and who to hire — then stay involved so the plan survives contact with reality.
More about this →
02Selected work
Three industries, three tailored systems.
A laboratory, a newsroom and a farm have nothing in common except that off-the-shelf software fell short. The pattern is the same each time: learn the domain, then build exactly what it needs.
01Laboratory medicine
Laboratory information system for molecular pathology
The pathology laboratory of a US health system
- Problem
- Molecular and cytopathology cases moved through a dozen steps — incoming sample, slide review, analysis, interpretation — with no single place to see where each one stood, and no way to look at the laboratory’s own data.
- What I built
- Case workflow tracker: status board, task queues and a per-test pipeline from sample to signed interpretation
- Automation of the processing steps that used to be done by hand, with barcode-driven lookup
- Analytics dashboards: diagnostic category distributions, mutation counts, call volumes over time, patient demographics
- Real-time waiting-time monitor for the sample collection area
- Outcome
- In daily production use by the laboratory staff, shipped in dozens of releases over several years — the system the rest of my work grew around.
02Journalism
AI news monitor for investigative journalism
An investigative newsroom covering organized crime across Latin America
- Problem
- Reporters needed to collect, filter, classify and search thousands of articles from hundreds of regional publishers — far more than a team can read.
- What I built
- Crawling, scraping and indexing pipeline over 250+ publishers, with per-source health monitoring
- NLP classification, entity recognition and a recommender that ranks what each reporter should read
- Semantic search and a chatbot with retrieval-augmented generation over a large vector database
- Outcome
- 12× more articles covered, higher classification precision, and natural-language access to the whole archive.
03Agriculture
Planning and logistics for large-scale agriculture
A large-scale flower grower
- Problem
- Daily field work — harvesting, sorting, packing, cultural labors — had to be planned across dozens of blocks, crews and supervisors, and production had to be predicted weeks ahead.
- What I built
- Task planning by day and week: crews, supervisors and blocks on one agenda and one map
- Route optimization and constrained resource planning for the field operation
- Crop-cycle predictor: expected stems per product and date, for precision agriculture
- Performance indicators for the operation, from the office to the field
- Outcome
- One system for planning, dispatch and prediction, running the operation of a farm with dozens of production blocks.
03About
Engineer first, architect by necessity.
I started as a full-stack developer and never stopped writing code. The architecture part came from watching good products slow down under their own weight — and deciding that the foundations deserved as much care as the features.
Curiosity took me through very different industries; the deepest roots are in clinical laboratories inside US health systems, where software has to be right every time. That range, plus covering the whole stack myself, is what lets me deliver tailored solutions instead of fitting your problem to a package I resell: service as a software, not software as a service.
Today I co-lead Impossible Code, an AI software company operating from Medellín with international clients, where I run the development team as Chief Architect and still ship code every week.
- Based in
- Medellín, Colombia
- Working with
- International clients, remotely
- Languages
- Spanish, English
- Company
- Impossible Code
04Questions
Things people ask before they write.
What kind of work do you take on?
Three things: building AI-powered products end to end, designing (or rescuing) software architecture and infrastructure, and technology consulting — including fractional CTO work. If your problem sits somewhere between those, it is probably a good fit.
Do you work with international clients?
Yes — most of my work is international and fully remote. I am based in Medellín and work in GMT-5, which shares the working day with the Americas and overlaps European mornings. Everything happens in English or Spanish, whichever the team prefers.
Do you sell a product, a platform or a license?
No. I sell service as a software rather than software as a service: senior engineering applied to your specific problem, delivered as working systems that you own. There is no package to license and no vendor to stay locked into — the solution is tailored to how your business actually works.
How does an engagement start?
Send me a WhatsApp message with a few lines about what you are building and where it hurts. We set up a call, and if there is a fit you get a short written proposal with scope, timeline and price.