I build AI agents, automation, and software, usually for finance and ops teams.
Penn State MBA finishing May 2026. Founder of AutomatifAI. Based in Santiago, Chile. Open to remote work for companies anywhere.
📧 [email protected] · 📞 +56 9 7809 5762 · LinkedIn · GitHub
I started as a structural engineer. I wrote C++ code for OpenSees during my master's thesis (a finite element framework used in seismic research) and worked remotely for an Italian software company building plugins for it. Around 2023 I got curious about RPA and joined a Chilean automation consultancy called Sisua Digital, where I spent a year deploying UiPath bots and Python scripts for finance teams across LatAm.
Somewhere in there I realized I'd rather build the tools than execute the workflows. I started AutomatifAI in 2024, a SaaS that ships AI agents on top of WhatsApp and Telegram for customer service teams. Eight enterprise clients later, it's still running, profitable, and growing slowly. I run it as a side project while I finish my MBA at Penn State.
The MBA was a deliberate move. I had the engineering and code background. I was missing the financial reasoning that would let me build software for finance specifically. I interned at a Chilean boutique IB last year (Kaiross) and learned how M&A and DCF actually work in practice, not just from a textbook. Currently I'm working with two Smeal professors on AI-native financial modeling, funded by a school grant.
That's the through-line: engineering, then automation, then AI, then finance. None of it was planned. It mostly happened because I kept following the parts that were interesting.
The Smeal AI Grant projects. Three open-source tools for finance education. The most useful one so far is FinModel. It pulls 10-K data from SEC EDGAR and generates DCF and LBO models you can audit line by line. Used by some MBA students at Smeal as a starting point for their valuation work. The other two (DealScope, MarketPulse) are still in beta. All three are at github.com/j0selarenas.
AutomatifAI. Slow growth phase. Eight clients, all retained, mostly Chilean mid-market. I take new clients selectively because the deployment work still requires my time. Working on getting that down so I can take more.
Job search. Looking for a full-time role for after I graduate (May 2026). Mostly Chilean tech and fintech, some remote roles for US/regional companies. More on this below.
What it is: a SaaS that puts AI agents on top of WhatsApp and Telegram for customer service teams. Most chatbot products either feel robotic or break when the conversation gets weird. We built ours to handle real conversational edge cases without falling over.
The interesting technical bit: we route different types of requests to different LLMs (GPT, Claude, Gemini) based on cost vs accuracy tradeoffs. A simple FAQ lookup goes to Gemini Flash. A nuanced refund negotiation goes to Claude. This cut our inference costs by about 40% without dropping quality. The whole thing runs on Python with Redis caching, low-latency, around 5,000 requests a day.
The hardest part wasn't the engineering. It was getting clients to describe their workflows accurately. Most of them couldn't tell me how their support team actually handled tickets, only how it was supposed to work. So I built deployment around a structured discovery process that pulls the real workflow out of them in 4-6 hours of meetings. Once that's done, the bot itself is mostly templating.