Software Engineer · Applied AI

Felipe
Velasquez

Building with Intention

Researching open generative AI models and developing tools that hold up in real use. Founder of EDGE, a startup betting that AI search shouldn't cost you your data.

See my work Get in touch
Foster City · CA Scroll
Who I am

Engineer by training,
skeptic by practice.

Software engineer building GenAI systems. Backend, frontend, and the AI layer that ties them together, from raw idea to something people actually use.

I've worked across AI automation, IT infrastructure, and web development — hands-on roles where you own the whole thing end to end.

The part I've come to care about most is evaluation — measuring what a model actually does instead of what it appears to do. It's what decides whether anything ships.

Education

B.S. in Computer Science

Tusculum University · Greeneville, TN

Graduated · May 2025
0 GPA
★ Magna Cum Laude

Academic & Athletic Scholarship Recipient

Coursework

Data Structures & Algorithms · Machine Learning · Operating Systems · Software Engineering · Database Management · Cybersecurity

The work

Experience

Where I've owned real responsibility — the product, the systems behind it, and the infrastructure underneath.

Aug 2025 – Present
Founder & Lead Engineer
EDGE · Foster City, CA

A local-first AI search engine for businesses. All retrieval and inference run on-device, with zero cloud. I designed the full stack — ingestion, indexing, a hybrid semantic + keyword retrieval engine tuned for low-latency search over unstructured business data, and an on-device LLM inference service. Alongside it I built the evaluation harness, validating answers against ground truth across multiple query types. Systematic evaluation of local open-source models (Llama and Qwen via Ollama) set the routing: lighter tasks go to the smaller, faster model with no accuracy loss, while the larger is reserved for answer synthesis. Lead engineer on a small founding team, shipping across a Python backend, Next.js frontend, and Linux service layer, owning the full deploy/debug lifecycle.

Generative AIRetrieval-Augmented Generation (RAG)PythonFull-Stack DevelopmentLarge Language Models (LLM)
Sep 2025 – Jun 2026 · Part-time
IT Systems & AI Automation Specialist
TechPro Builders · San Carlos, CA

Sole owner of IT and AI automation for a construction company. I automated critical business workflows — expense tracking, inventory, and reporting — cutting hours of manual work each week along with operating costs, and prototyped LLM-powered tools on a local Linux workstation for business and field-technician workflows. I owned the company's full technical infrastructure end to end: storage and backup, networking, security, and broadband. I also identified a service gap and led the launch of a recurring-revenue program, designing the model and driving the rollout with the team.

AI AutomationIT Infrastructure ManagementWorkflow AutomationBusiness Process AutomationProblem Solving
May – Aug 2024 · May 2025 – Present
Technology & Marketing — Intern → Staff
Evolution Burlingame / Peninsula Tennis Club · Burlingame, CA

What began as a summer internship turned into a role I kept. I look after the databases and digital systems that keep the club running — automating email-marketing workflows and rebuilding audience segmentation to improve outreach — and I maintain and debug the company website. I coach tennis alongside the technical work, drawing on years of disciplined training in the sport.

Marketing AutomationWeb DevelopmentDatabase Management System (DBMS)WordPressEmail Marketing
Selected work

Projects

Things I've built on my own terms — research, a full-stack platform, and a publication.

assets/ai_research/1.jpg
landscape · ~1200×800px
AI Efficiency Research
01 — Research

LLM Performance Benchmarking Study

Tusculum University · Jan – May 2025

A controlled evaluation framework comparing ChatGPT, Claude, DeepSeek, and Gemini across standardized programming tasks. I helped design the scoring criteria and measured output quality, accuracy, and efficiency across CS coursework and real-world software tasks, then analyzed where each model held up and where it didn't. Presented at a formal academic poster session.

Large Language Models (LLM)AI/ML EvaluationBenchmarking MethodologyPrompt EngineeringData Analysis
Boone Lake platform Boone Lake platform Boone Lake platform Boone Lake platform Boone Lake platform
02 — Full-stack

Boone Lake Water Testing Platform

Tusculum University · Jan – May 2025

A PHP/MySQL platform that helps a local environmental association monitor Boone Lake's water quality across multiple zones. I designed the database schema, built role-scoped access for staff and administrators, and added automated threshold alerts for contaminant levels, plus a data-visualization dashboard the team used to pull figures for their weekly reports.

PHPMySQLDatabase DesignRole-Based AccessFull-Stack Development
Toolbox

Technical skills

01 AI & Retrieval Systems

On-device inferenceHybrid search (FAISS + BM25)Evaluation & benchmarkingRAG pipelinesModel routing

02 Software & Data

PythonData structures & algorithmsDatabase schema designREST APIsMySQLNext.js

03 Systems & Operations

Linux / UbuntuCross-layer debuggingPerformance profilingGitWorkflow automation
Off the clock

A little more about me

I grew up in Bogotá, played NCAA Division II tennis in Tennessee, and now build in the Bay Area. Computers have had me since I was a kid, and video games never quite let go — but if there's daylight and something to play outside, that's where I'll be. Fan of every sport going, though tennis is the one I owe the most to.

The court

Tennis, then & now

Tennis has been there my whole life — training that still keeps me focused and sharp.

Elevation

Trails & the outdoors

Hiking + anything that gets me up and moving.

Motion

The open road

Distance running. I love the feeling of a challenge, and the drive to somewhere new.

Sound

Live electronic music

It changed my life. Endless supply of dopamine, nothing beats a good drop.

Let's talk

Let's build something
on the edge.

I'm looking for software engineering and applied-AI roles. I'm also happy to talk about retrieval, evaluation, or what it takes to run a model privately — write to me either way.

✉ fevelasquin@gmail.com in · LinkedIn ↓ Resume