Hi, I'm Agustin Biani
I take ambiguous problems and ship them as AI products that work in production.
AI Engineer & builder, with a product and delivery background.
Building with LLMs, RAG, and agents: owning the work from problem definition to launch, and leading delivery along the way. Fifteen-plus years shipping for international clients, mostly in the USA.
I'm an AI engineer and builder who takes products from problem definition to production, designing and shipping systems with LLMs, RAG, agents, and traditional software. I've built complete products end to end, like MigraLert (a machine-learning system that predicts migraine attacks from smartwatch biometric data, presented at the Latin American Congress of Headache) and Cefalia (a cross-platform app live in three languages).
What makes me different is the 15+ years behind it: I've led cross-functional teams and delivered complex digital products for international clients, mostly in the USA, as the direct point of contact with founders and stakeholders. So I don't just implement — I own scope, make tradeoffs across quality, cost, and reliability, communicate clearly with technical and nontechnical people, and ship.
Since 2022 I've focused that experience entirely on AI: custom AI tools, automation pipelines, LLM-powered products, and intelligent agents. I combine hands-on building with strong product judgment, Agile/Scrum leadership, and a relentless focus on delivery quality.
Stack & tools
Fifteen-plus years in technology services and consulting, acting as the primary point of contact for U.S. and international clients across multi-team engagements — and, since 2022, building the AI products those clients ship: internal RAG systems, fine-tuned LLMs for medical research, revenue and capacity forecasting models, and agents that replaced hours of daily work per person.
I build AI products with my own hands and lead the teams that ship them, from problem definition to production.
Internal RAG systems, LLMs fine-tuned for medical research, LLM-based external products and AI agents that take over repetitive work. Python, PyTorch and TensorFlow, with a linguistics background that helps when the problem is language.
Time-series forecasting for subscription revenue and project hours; biometric signal analysis and crisis prediction for MigraLert. End-to-end pipelines: ingestion, cloud databases (BigQuery, SQL), Streamlit dashboards and models in production.
From problem definition to launch: cross-platform mobile apps with React Native, web with JavaScript and HTML, automation pipelines and custom AI tooling. Two products shipped 0→1 and live today.
Agile/Scrum leadership of cross-functional teams, direct point of contact with founders and stakeholders, and ownership of scope, budget, deadlines and final quality. Tradeoffs across quality, cost and reliability, explained clearly to technical and nontechnical people.
Whether it's an internal tool, an LLM-powered product or a research pipeline, the shape is the same: define the problem properly, build the simplest thing that works, ship it, and keep delivery inside what was agreed.
1. Define
What decision or workflow does this change? What data exists? What does “good enough” look like? Milestones, owners and success metrics agreed before anything is built.
2. Build
A working RAG, agent or model in days, evaluated against real data; then the unglamorous part: reliability, cost, latency, evals and the integrations that make it usable.
3. Ship & measure
Launch, instrument, and measure against the metric agreed in step 1. Retrospectives that produce concrete changes for the next iteration, not just notes.
4. Lead the delivery
Remove blockers, flag risks early, and keep technical and nontechnical people on the same page. Status and reporting run on automations I build myself, so my attention stays on the product.
AI and health-tech products I have taken from problem definition to production — MigraLert and Cefalia designed and built end to end, solo.
Shipped cross-platform mobile app · React Native
The first fully free Latin American migraine & headache tracker, with local-first storage. Patients log attacks, triggers and treatments in seconds and get a clear history to share with their neurologist. Live on iOS and Android in Spanish, Portuguese and English (2026). Designed, built and shipped solo.
cefalia.app
Machine learning & data-science research · Python, Streamlit, BigQuery
A system that predicts migraine attacks from smartwatch biometric data. Full pipeline: watch app → cloud ingestion → cloud database → Python/Streamlit research dashboard → ML model that detects crises, identifies external triggers and crisis types, and forecasts upcoming episodes. Designed and built end to end, solo.
Presented at the Latin American Congress of Headache. Interim results: 67% prediction accuracy, targeting 80%+.
migralert.com
AI diagnostic tool built on machine learning
An AI tool that detects stroke symptoms through a standardized neurological test. First of its kind in Latin America, combining several ML algorithms.
neuroalerta.com
AI nurse for daily neuro-monitoring
An AI nurse for neuro-monitoring. A guided 5-minute daily test tracks progress and helps prevent decline, reporting results to the attending physician.
Have an AI product to build, a pipeline to put in production, or a team that needs someone who can both engineer and lead delivery? I'd love to hear from you. Open to remote roles and EU relocation.
agustin@biani.org
