AI Engineer · MSc in Artificial Intelligence
Frontend + AI

AI Engineer with hands-on delivery across LLM integration, NLP, and Computer Vision.

• Built an LLM-powered contract analysis system that cut review time by ~65% and reduced token consumption by 85–90% through prompt compression and context window optimization — deployed in production at Transgenia (Mexico).
• Developed the Film Tax Credit Agent, a Python-based agentic system that automates eligibility and credit calculation for Dominican audiovisual productions under Art. 39 — replacing a fully manual compliance process.
• Completed an MSc in Artificial Intelligence (UNIR, 2026) with a Master's Thesis on hybrid NLP recommendation systems, achieving 0.815 avg similarity vs 0.13 baseline across 12 test queries.
• 5+ years of financial data analysis domain expertise in the Dominican public sector — directly applied to AI system design.

I build production-grade AI, not prototypes.

AI Agents · Python · Business Rules Automation

Python-based agentic system that automates eligibility evaluation and tax credit calculation for Dominican audiovisual productions under Art. 39. Replaces a fully manual compliance process with end-to-end tested business logic covering validated expenditure calculation, credit determination, and audit traceability. Designed to scale batch processing at near-zero marginal cost.

🔗 https://github.com/jorgedoiany/film-tax-credit-agent

NLP · Hybrid Recommendation System · Streamlit

Hybrid NLP recommendation system that matches narrative scene descriptions to real film locations in the Dominican Republic. Combines TF-IDF with MiniLM and XLM-R SBERT sentence embeddings using cosine similarity. Trained on a real dataset of 208 verified locations provided by DGCINE. Hybrid model achieved 0.815 avg similarity vs 0.13 TF-IDF baseline across 12 test queries.

Master's Thesis — MSc in Artificial Intelligence, UNIR (2026)

🔗 https://github.com/jorgedoiany/film-location-recommender-pilot

Computer Vision · Temporal Analysis · OpenCV

Computer vision system that detects and quantifies deforestation in the Amazon using satellite image time series (2000–2019). Compares 3 detection algorithms: Baseline Gradient, HSV Vegetation Loss, and Robust Multi-Space (HSV + LAB + cloud filtering). Includes CLI batch processing, Jupyter notebooks, and pytest-validated results.

🔗 https://github.com/jorgedoiany/amazon-deforestation-segmentation

Computer Vision · Image Processing · Evaluation Metrics

Reproducible pipeline comparing classical low-light enhancement methods: Gamma Correction, CLAHE, and Single-Scale Retinex (SSR). Quantitative evaluation using PSNR and SSIM metrics. CLI execution, Jupyter demo notebook, and structured CSV output per run.

🔗 https://github.com/jorgedoiany/low-light-image-enhancement

📌 Master’s Thesis (TFM):
AI-based hybrid recommender system for film location scouting
(Currently in development — experimental and applied focus)

• LLM Integration · Prompt Engineering · Context Window Optimization
• NLP: TF-IDF · Sentence Embeddings · Cosine Similarity · spaCy
• Computer Vision: Image Segmentation · Enhancement Pipelines · Temporal Analysis
• Recommendation Systems: Hybrid Models · Semantic Search
• Model evaluation · Reproducible experimentation

Frontend background applied to AI-driven interfaces and data products.

_{Last updated: 2026}