Jancarlos Sosa

Selected work, field notes, and a direct line.

Profile / 01

Engineer by training.
Systems thinker by habit.

I’m an AI engineer building at the intersection of models, software, and operational reality.

Jancarlos Sosa wearing a gray suit and striped tie
Jancarlos SosaAI Engineer

How I work / 02

Useful before impressive.

I prefer a clear workflow with measurable value over a clever demonstration looking for a problem. That means understanding the human process before selecting the model.

Control is a feature.

Agentic systems should expose consequential decisions, require approval where appropriate, and leave enough evidence to understand what happened.

Full stack means full context.

Reliable AI depends on more than inference: data models, APIs, authentication, interfaces, deployment, monitoring, and the quality of every boundary between them.

Experience / 03

AI Engineer

Infosys

Designing agentic systems for enterprise workflows where perception, automation, and human review must work together reliably.

  • Built a computer-use testing system for a legacy desktop application using multimodal perception, OCR, and parallel execution.
  • Created a human-approved remediation pipeline connecting issue intake, coding agents, pull requests, and closure tracking.
  • Prototyped retrieval workflows with Python, LangChain, and vector search to turn support signals into development priorities.

Python Flask Developer

University of Connecticut

Developed database-backed workflows and interfaces for a university research group.

  • Connected Flask services to MySQL and evolved schemas as research workflows changed.
  • Built interfaces for profiles, activity records, projects, and collaboration data.

InStep Technology Intern

Infosys

Built a full-stack application with a small engineering team using Spring Boot, Angular, and MySQL.

  • Designed REST endpoints, data models, and responsive application views.
  • Collaborated across backend, frontend, and database responsibilities.

Working range / 04

Agent systems

MCP, computer use, tool orchestration, human-in-the-loop controls

Applied AI

RAG, multimodal models, OCR, speech, evaluation-minded workflows

Product engineering

Python, Java, TypeScript, FastAPI, Spring Boot, React, Next.js

Cloud & data

AWS, PostgreSQL, Supabase, Docker, APIs, authentication

Foundation / 05

B.S. Computer Science

University of Connecticut · 2025
Concentration in Software Design and Development