Computer Science Engineering student at UTPL, Ecuador. I work on spiking neural networks built from real connectome data.
LinkedIn · Email · ORCID · Loja, Ecuador
In October 2024 the FlyWire consortium published the first synapse-resolution map of an adult Drosophila brain: about 128,000 neurons and 53 million synapses. Shiu et al. then showed that this wiring alone, simulated as a leaky integrate-and-fire network, predicts real fly motor behaviour at roughly 95% accuracy. In March 2026 Eon Systems ran a whole embodied emulation of the fly in a physics engine.
All of those models share one limitation, and their authors say so themselves: the synaptic weights are static. The networks don't learn.
My undergraduate thesis adds the learning. I extract the Johnston's Organ auditory circuit from the FlyWire connectome, implement it in Brian2, add spike-timing-dependent plasticity and winner-take-all dynamics, and race it against a control network matched on neuron count, degree distribution and everything else I could hold fixed. Topology is the only thing that differs, so if one wins, I know why.
The question underneath it: does a topology shaped by evolution actually compute better than one we design? Nobody could test that before 2024.
→ Code and experiments: flywire-snn-stdp
AI Evaluation Specialist / Technical Data Annotator — LinkedIn, contract via GreenLight Workforce Solutions, since Apr 2026 Benchmark datasets and structured annotations for code generation and code editing models. Most of my time goes into finding the specific places where models break on multi-step, realistic engineering tasks.
AI Research & Evaluation Engineer at Scale AI, Jul 2025 – Jul 2026 LLM evaluation pipelines covering reasoning, safety, factuality, tool use and MCP compliance, with Python statistical logic as the ground-truth validator. RLHF and RLAIF evaluation, adversarial prompting, schema-based evaluation frameworks.
Machine Learning Engineer at Clipp, part-time, Oct 2024 – Aug 2025 Analytics across six product lines. Clustering on raw event logs to segment user behaviour, plus the feature engineering that made those models work at all.
Full Stack Developer at Clínica San José, Oct 2024 – Feb 2025 Clinic management system in React and Node, deployed and used in production.
flywire-snn-stdp — Connectome-derived spiking neural network with STDP plasticity, benchmarked against a degree-matched random control. Brian2, NetworkX, Python.
Clipp Analytics Platform — Microservice platform producing KPI dashboards for six Clipp service lines. I designed the architecture and built the ETL module: scheduled Airflow extraction into a central PostgreSQL store, served through FastAPI. Team of three. Docker, Airflow, FastAPI, Pandas, PostgreSQL, Next.js.
SIGMA-IA Clipp — User segmentation for the Clipp Events app, so advertising targets preference and behaviour instead of broadcasting. I led the project and built the clustering models. Team of three. Python, scikit-learn.
Insurance Policy AI Assistant — Multi-tool agent for insurance workflows. Team project; I built the agent core and the retrieval pipeline: Gemini integration, the FAISS-backed RAG layer, chat history and UI, source-citation metadata on every response, and a web-search tool for anything the index doesn't hold. I measured retrieval quality with RAGAS instead of judging answers by eye. LangChain, FAISS, Gemini API, RAGAS, FastAPI, Docker.
Garantec Quoting Platform — Replacing a manual vehicle-insurance quoting process that took two to three days with one that targets under three minutes. I worked as project manager. Team of four. React, Material UI, Symfony REST backend.
Python, Brian2, NumPy, Pandas, NetworkX, scikit-learn, PyTorch, TensorFlow, LangChain JavaScript, React, Node.js, Express, SQL, FastAPI, Docker, AWS, Git
Computer Science Engineering, Universidad Técnica Particular de Loja, 2023 – expected Mar 2027 Thesis supervisor: Prof. Alexandra Cristina Gonzáles Eras
AnyoneAI, Machine Learning Engineer Specialization, 2025. Project-based industry programme, not a degree. Covered credit-risk modelling with gradient boosting, CNN image classification deployed on AWS, and ELT pipelines.
I'm looking for a research internship or research assistant position in neuromorphic computing or computational neuroscience, in Europe or elsewhere. If you work on spiking networks, connectomics or energy-efficient AI and something here overlaps with what you're doing, I'd like to hear from you: ljmora004@outlook.com