# Victor F. Ayllon

**AI/ML Engineer**

Chicago, IL · vayllon301@gmail.com · +1 312 481 1538 · [vayllon.me](https://vayllon.me) · [linkedin.com/in/victor-ayllon](https://www.linkedin.com/in/victor-ayllon-fernandez-79723b244/) · [github.com/vayllon301](https://github.com/vayllon301)

## Summary

I'm a results-oriented AI/ML Engineer who enjoys building things and exploring new technologies as they land. I like carrying a problem from idea to production. That has spanned autonomous agents for security operations at Boehringer Ingelheim, LangGraph agents and FastAPI services behind Next.js and React Native clients at Konnectia, and Junsu, the AI compliance platform I co-founded, which already generates revenue from enterprise customers - Graph RAG retrieval with audit-grade citation provenance, human-in-the-loop review, and token-level cost control.

## Professional Experience

### AI Security Engineer

**Boehringer Ingelheim · Enterprise security operations | Mar 2026 - Present**

- Co-designed and implemented the Triage Agent, which automates the cybersecurity triage layer by investigating incidents across Microsoft Sentinel and other security tooling, evaluating findings, and autonomously resolving or escalating on severity.
- Created a UI to visualize access policies and group inclusions and exclusions, so security teams can inspect entitlement structures and exception rules without digging through raw policy configurations.
- Analyzed a 70,000-incident security dataset to surface operational performance weak spots, and built the interactive platform presenting those findings to the security team.

### Co-Founder

**Junsu · AI compliance platform · junsu.es | Feb 2025 - Present**

- Built the pgvector retrieval layer over customer compliance evidence, with page and character-offset provenance so every AI recommendation cites a source span an auditor can verify.
- Designed the LLM assessment pipeline mapping controls to risks: one batched call per activity to keep related judgments consistent, the rubric cached as a prompt prefix to cut cost, and output excluded from scoring until a human confirms it.
- Shipped on AWS (Bedrock, S3, Cognito, Lambda) with per-tenant Postgres row-level security and a token-level cost ledger reconciled against the AWS bill. Revenue-generating with enterprise customers, covering ISO 27001, UNE 19601, ESG and B Corp.

### AI Engineer

**Konnectia · Healthcare AI product team | Sep 2025 - Sep 2026**

- Owned MenteViva end to end, an AI agent for elderly care that continuously evaluates user status, generates alerts, and delivers personalized recommendations.
- Architected the Graph RAG memory layer that stores and retrieves contextual data about the elder's family, friends, and memories, grounding every agent response in personal history.
- Developed a custom routing and guardrail layer with LiteLLM.
- Shipped and maintained the full delivery surface: LangGraph orchestration and FastAPI services behind a Next.js web frontend and a React Native mobile application.

### Teaching Assistant, Programming Methodology and Technology

**La Salle URL | Sep 2024 - Jul 2025**

- Mentored students through academic projects, resolving technical blockers and programming questions.

## Education

### Data Science M.S.

**Lewis University | August 2026 - July 2027**

Final year of La Salle's international dual-degree track, earning the M.S. alongside the engineering degree.

### Exchange Program - Cybersecurity & Data Science

**Fu Jen Catholic University, Taipei · GPA 4.0 | August 2025 - February 2026**

Coursework in cybersecurity, business intelligence and data science.

### International Computer Engineering

**La Salle - URL | September 2022 - July 2027**

Five-year international track concluding with a Data Science M.S. at Lewis University (USA).

## Technical Skills

- **Agentic AI & LLM Systems:** RAG, Graph RAG, ontology-based retrieval, multi-agent systems, MCP, LangGraph, LlamaIndex, LiteLLM, LoRA/PEFT fine-tuning
- **Agentic Coding Tools:** Claude Code, Cursor, Codex (daily development workflow), MCP server authoring
- **Machine Learning / Data Science:** scikit-learn, pandas, NumPy, PyTorch, Databricks, MLflow
- **Data & Retrieval:** PostgreSQL, Redis, Neo4j, pgvector
- **Web Engineering:** FastAPI, React, Next.js, React Native, Tailwind CSS
- **Cloud & Platforms:** Azure, AWS, Supabase, Vercel
- **Languages:** Spanish (Native), Catalan (Native), English (C2), French (Basic)

## Projects

### MenteViva — AI agent for elderly care

https://menteviva.konnectia.net

AI product for continuous elder status evaluation, alert generation, and personalized recommendations.

- Graph RAG memory layer for family, friends, and personal context.
- Next.js web frontend and React Native mobile application.
- Designed for personalized, context-aware interaction.
- **Stack:** LangGraph, Graph RAG, Next.js, React Native, FastAPI

### Junsu — AI Compliance Platform

https://junsu.es

AI-driven platform ensuring regulatory compliance through automated document analysis and policy rule extraction.

- Created a custom tool for compliance assessments powered by AI recommendations based on assessed company investigation and data provided.
- Full AWS platform with compliance at the core of the product, SOC II and ISO 27001.
- **Stack:** React, AWS (Bedrock, S3, Amplify...), Graph RAG