Enterprise-grade AI systems for the pharmaceutical sector: LLM pipelines, multi-agent orchestration and MCP-based integrations on AWS, delivered across full DEV → QA → PROD lifecycles.
7+ years translating complex information into high-impact decisions across research (European Institute of Consciousness Research · The Beckley Foundation), international organizations (UN/UNDP — 2030 Agenda) and regulated industry.
flowchart TB
subgraph cicd["⚙️ Delivery pipeline — 3 environments"]
direction LR
git["📦 Git · Bitbucket"] -->|PR + code review| dev["🔧 DEV<br/>Docker · feature"]
dev -->|integration tests| qa["🧪 QA<br/>Docker · release"]
qa -->|approval gate| prod["🚀 PROD<br/>Docker · main"]
end
subgraph aws["☁️ AWS"]
direction LR
ec2["🖥️ EC2<br/>compute · containers"] --- rds[("🗄️ RDS<br/>relational data")]
etl["🔄 ETL jobs<br/>scheduled · event-driven"] --> rds
end
subgraph integration["🔌 Integration layer"]
direction LR
oauth["🔐 OAuth 2.0<br/>M365 · Graph · device-code"] --> mcp["⚙️ MCP servers"]
mcp --> skills["🧩 Skills system<br/>query · transform · report"]
skills --> scoping["🛡️ Per-role<br/>data scoping"]
end
subgraph ai["🤖 AI layer"]
direction LR
agents["🕸️ LangGraph<br/>multi-agent orchestration"] --- rag["📚 RAG<br/>knowledge synthesis"]
rag --- stt["🎙️ Audio → text<br/>clinical transcription"]
guard["✅ Validation gates<br/>deterministic post-checks"]
end
subgraph products["📤 Products"]
direction LR
radars["🧬 Molecular radars<br/>automated surveillance"]
reports["📋 Scientific reports<br/>standards-gated"]
prompts["⚡ Programmable prompts<br/>analytics automation"]
end
sources["🌐 Molecular & clinical<br/>data sources"] --> etl
prod -->|deploys to| ec2
ec2 --> mcp
rds --> rag
scoping --> agents
agents --> guard
guard --> radars & reports & prompts
radars -->|criteria match| hooks["🪝 Automatic hooks<br/>alerts · downstream triggers"]
reports --> users["👥 Reviewers & analysts"]
prompts --> users
hooks --> users
Production code is private (regulated industry). What it does:
| Domain | Project | Description |
|---|---|---|
| Data Pipeline | Molecular search radars | Scheduled radars sweep molecular and clinical data sources; automatic hooks trigger downstream analysis and alerting on criteria matches. Python · AWS EC2/RDS · event-driven across DEV/QA/PROD |
| LLM and RAG | Scientific report generation | LLM pipelines drafting pharmacological research reports under established scientific criteria — structured extraction, citation discipline, validation gates. RAG + LangGraph with deterministic post-checks |
| AI Platform | MCP servers + Skills system | Model Context Protocol servers exposing enterprise data/tooling as composable skills — safe, scoped capabilities (query, transform, report) with OAuth and per-role data scoping |
| Analytics | AI enablers for analytics | Bridges giving LLMs access to enterprise analytics software, returning structured programmable prompts — copy-ready artifacts that reproduce and automate analyses |
| Speech AI | Audio-to-text agents | Clinical transcription pipelines: speech → structured medical documentation with terminology normalization |
| Security | OAuth & identity | Device-code and delegated flows for Microsoft 365 / Graph; token lifecycle for headless agents |
- microsoftgraph/msgraph-sdk-python-core#1089 — cut ~20-25% of
GraphServiceClientcold-import time by deferring an undeclaredrequestsimport (profiled withpython -X importtime) - msgraph-sdk-python#904 — import-time profiling and startup workaround for the SDK
- Positron discussions — root-cause analysis: interpreter discovery in containerized pixi environments
Engineering practice — 3-environment delivery (DEV/QA/PROD) · Git (GitHub/Bitbucket) · OOP/functional · code review · OAuth2 · CLI-first workflow (WSL2 · tmux · nvim)
Statistics & economics — Bayesian inference · multivariate · time series · econometrics · Monte Carlo · forecasting · spatial analysis
- Engineering in AI (foundational program) — TecNM, 2025–2026 · Applied AI, ML/DL/RL, Generative & Agentic AI
- M.Sc. Data Science, Statistics & Probability — UCJC Madrid (EU-recognized, research focus) · Thesis: ML for consciousness-state classification from multimodal neuroimaging
- B.Sc. Economics — top 5% nationally (CENEVAL EGEL-ECO)
- 24 professional certifications — IBM (ML with Python, Honors) · DeepLearning.AI · Microsoft · DataCamp · Meta · Packt
- Collaborations — UN/UNDP (official SDG report, State of Mexico) · European Institute of Consciousness Research · The Beckley Foundation · INAWE Observatory