Aregnys C.Paredes Gomez

What I focus on

• Generative AI & LLMs

  Integrating commercial and open-source LLMs into production workflows with retrieval-augmented generation and structured prompting.

  • #rag
  • #fine-tuning
  • #openai-api
  • #hugging-face

• Multi-Agent Systems

  Coordinating specialized agents for retrieval, extraction, validation, and summarization, with tool use, memory, and guardrails for reliable outputs.

  • #langchain
  • #tool-use
  • #guardrails

• Model Evaluation & Data Quality

  Golden datasets, scoring rubrics, and automated regression tests that track accuracy, hallucination, and consistency, plus labeling guidelines and reviewer calibration.

  • #benchmarking
  • #error-analysis
  • #annotation

• Document Intelligence

  Pipelines that combine OCR, layout analysis, and LLM-based extraction to turn contracts, invoices, and forms into validated structured data.

  • #ocr
  • #layout-analysis
  • #extraction

• MLOps

  Containerized model serving on AWS and Azure, CI/CD for models and prompts, and monitoring for latency, cost, and model drift.

  • #docker
  • #kubernetes
  • #ci/cd

• AI Security

  PII redaction, role-based access control, and defenses against prompt injection and data leakage, grounded in years of secure financial systems work.

  • #pii-redaction
  • #rbac
  • #prompt-injection-defense

1. Senior AI / Software Engineer

   Blackstraw AI · Tampa, FL

   Oct 2022 – Present

   Focus: Generative AI, multi-agent systems, MLOps, document intelligence

   • Architect and deploy generative AI solutions for enterprise clients, integrating commercial and open-source LLMs (OpenAI API, Hugging Face) into production workflows using retrieval-augmented generation and structured prompting.
   • Design multi-agent systems in LangChain that coordinate specialized agents for retrieval, extraction, validation, and summarization, with tool use, memory, and guardrails for reliable outputs.
   • Build document intelligence pipelines that combine OCR, layout analysis, and LLM-based extraction to convert unstructured documents such as contracts, invoices, and forms into validated structured data.
   • Fine-tune and evaluate LLMs on domain-specific datasets; create golden datasets, scoring rubrics, and automated regression tests to track accuracy, hallucination, and consistency across model and prompt versions.
   • Lead data annotation and quality workflows, including labeling guidelines, reviewer calibration, and error analysis that feed targeted improvements back into training and evaluation data.
   • Own MLOps for AI services: containerized model serving with Docker and Kubernetes on AWS and Azure, CI/CD for models and prompts, and monitoring for latency, cost, and model drift.
   • Apply security best practices to AI systems, including PII redaction, role-based access control, and defenses against prompt injection and data leakage.

2. Machine Learning Engineer

   NLP Logix · Jacksonville, FL

   Jul 2020 – Aug 2022

   Focus: Machine learning, NLP, computer vision, data science, AI engineering

   • Developed and deployed NLP models for text classification, named entity recognition, and sentiment analysis using PyTorch, TensorFlow, and Hugging Face Transformers.
   • Built computer vision models for image classification and object detection, handling dataset preparation, augmentation, training, and production deployment.
   • Engineered end-to-end data pipelines and feature engineering workflows in Python and Pandas, with validation checks to ensure clean, consistent training data.
   • Ran rigorous model evaluation using cross-validation, precision/recall and confusion-matrix analysis, and error analysis to guide model selection and retraining.
   • Served models through FastAPI and Django microservices and translated client business problems into measurable ML objectives alongside data science and product teams.

3. Full-Stack Developer

   Armada Labs · Estero, FL

   Mar 2016 – May 2020

   Focus: Financial software, lending, payments, digital banking

   • Built full-stack features for digital banking and lending platforms using Node.js, NestJS, TypeScript, and PostgreSQL/MongoDB.
   • Developed loan origination and servicing workflows, including application intake, underwriting rules, and repayment scheduling.
   • Integrated payment processing and designed secure REST APIs with OAuth2/JWT authentication and encryption of sensitive financial data in transit and at rest.
   • Implemented event-driven services with Kafka and RabbitMQ and Redis caching to improve reliability and responsiveness of transaction flows.
   • Performed code reviews focused on secure coding and remediated vulnerabilities identified in security assessments.

4. Software Developer

   EFX Financial Services · Largo, FL

   Feb 2012 – Jan 2016

   Focus: Transaction processing, proprietary financial software

   • Developed and maintained high-throughput transaction processing components in C/C++ and Java for proprietary financial software.
   • Built networked services over TCP/IP, including socket-level communication and message protocols between internal systems.
   • Optimized performance through multithreading, memory management, and SQL query and stored-procedure tuning.
   • Implemented security controls including encryption, authentication, and audit logging to protect financial transaction data.