MSc Artificial Intelligence (Distinction) — Queen Mary University of London
I am a Senior Data Scientist and AI/NLP Researcher based in London, UK, combining 5+ years of enterprise machine learning experience in Tier-1 banking with top-tier academic AI research.
- 🎓 Graduated with Distinction in MSc Artificial Intelligence (NLP Stream) at Queen Mary University of London (QMUL) and First Class Honours in Electronic Engineering at PUCP (Ranked 4th in graduating class).
- 💼 Former Lead / Advanced Data Scientist at BCP (largest bank in Peru) and BBVA, where I led experimentation platforms driving $1M+ incremental profit, engineered fraud susceptibility models (65% detection rate), and built cash remittance forecasting algorithms saving £110k+ annually.
- 🔬 Published NLP Researcher with papers at NAACL (SYGTIP) and AmericasNLP, winner of 1st Place in the AmericasNLP 2021 Shared Task on machine translation for Quechua, and author of the Awajun-OP open-access parallel dataset with Ojo Público.
- 🎯 Deep expertise in Parameter-Efficient Fine-Tuning (LoRA, PiSSA, DoRA), Continual Learning (CaLoRA), Multi-Tier Information Retrieval (MRL, Stella, FAISS), and Vision-Language / Multimodal Reasoning (LongCLIP, LLaVA).
| Repository | Domain | Core Tech & Architecture | Highlight Benchmark / Accolade |
|---|---|---|---|
| 🏛️ PEFT-CL_Spanish_Peruvian_NMT | MSc AI Thesis | Meta NLLB-200-1.3B, LoRA, PiSSA, FFT, CaLoRA continual learning |
MSc AI Thesis (Distinction) across 5 Peruvian low-resource languages |
| 🔍 scifact-tiered-neural-search | Neural Search & IR | Stella-1.5B, Qwen2.5-3B HyDE, GOLFer, MRL 256D, FAISS, Streamlit |
nDCG@10: 0.8063, Recall@10: 93.0% (Best Proposal & Top 2 System) |
| 📊 hybrid-tweet-sentiment-analysis | NLP & ML Ensembles | Multi-Lexicon (VADER, AFINN, Hu & Liu), SentenceTransformers, LightGBM |
Score: 97/100 (Cohort Rank #1), 91.04% CV F1 |
| 👁️ longclip-llava-multimodal-vqa | Multimodal Vision-Language | LongCLIP-ViT-B/32 (248 tokens), LLaVA, Qwen-0.6B, PiSSA PEFT |
67.62% Test VQA Accuracy (Rank #18 Cohort) |
| 🚗 risk-sensitive-autonomous-driving-rl | Deep RL & Safety | QR-DQN, Distributional RL, CVaR Safety Shield, Highway-Env |
Crash Rate: 13.8% (5.2x safer than DQN) |
| 🧩 visual-semantic-embeddings-and-planning | Neuro-Symbolic AI | Word2Vec Skip-Gram, PyTorch CIFAR-100 Projection, PDDL Planning |
mAP@K: 0.8079, 18/18 Plans Solved (Top 5% Cohort) |
- Findings of the AmericasNLP 2025 Shared Tasks on Machine Translation, Creation of Educational Material, and Translation Metrics for Indigenous Languages of the Americas.
The Fifth Workshop on NLP for Indigenous Languages of the Americas (May 2025). Paper - Awajun-OP: Multi-domain dataset for Spanish–Awajun Machine Translation.
Workshop on NLP for Indigenous Languages of the Americas (March 2024). Dataset Repo - Exploring Linguistic Typology Features in Multilingual Machine Translation.
SYGTIP Workshop at NAACL 2021. Paper - The REPU CS’ Spanish–Quechua submission to the AmericasNLP 2021 Shared Task on Open Machine Translation.
Proceedings of the First Workshop on NLP for Indigenous Languages of the Americas (June 2021). 🏆 1st Place Winner for Quechua (Track 2). Paper