Skip to content

Alexis Roldan

Full-Stack Data Scientist & Sr. Data Engineer

Lancaster, CA. alexis.m.roldan.ds@gmail.com

Baxter 2005–2015, Baxalta 2015–2016, Shire 2016–2019, Takeda 2019–Present

PDF version

Full-Stack Data Scientist & Sr. Data Engineer at Takeda Pharmaceutical, on a global AI and data science team for plasma manufacturing. I build manufacturing analytics and GenAI/RAG systems end to end — from ETL and ML modeling to LLM apps deployed in production. Twenty years in biopharma, thirteen in data roles.

Experience

  1. Aug 2026 – Present

    Sr. Data Engineer & Data Scientist (Sr. Manager)

    Takeda – Global

    Global AI / Data Science team (GMS) building GenAI and analytics solutions for Takeda's plasma manufacturing sites; still own AMIRA and BioTrack.

    • Moved in Aug 2026 from the LA site team into a newly formed global AI / Data Science team for Takeda's plasma manufacturing network — solutions configured per site, one engine, instead of one-site tools
    • Bring the patterns the team's multi-site solutions build on, proven on two production platforms: R computes / LLM narrates, configuration per site with one engine, the Unity Catalog data layer, usage instrumented from day one, and the 16-artifact SDLC dossier that closed a platform audit finding
    • Authored the proposal and built the proof of concept (dummy data) for a lab-investigation closure tool, the team's first project through the SDLC
    • Still own AMIRA (production, v1.10.2: per-session freshness, 322 test blocks, an evidence line under every TOMO answer) and BioTrack (production, v1.2.1 'the transparency release': one alert-rate function at ~40 call sites, 1,065 assertions, 0 failures)
    • Primary stack R/Shiny on the Databricks engine; React/Node.js where it fits

  2. Apr 2025 – Aug 2026

    Sr. Data Engineer & Data Scientist (Sr. Manager)

    Takeda Pharmaceutical

    Supported the Los Angeles Digital Roadmap, AI/ML development and advanced analytics to drive data-driven decisions.

    • Owned the LA site's AI and analytics platform portfolio end to end — ETL, modeling, app, deployment: AMIRA and BioTrack in production, Horizon and LabCast as proofs of concept, plus the team's solutions portal; AMIRA and its TOMO assistant carry an initial estimate of 4,000+ hours returned through near-real-time analytics and data aggregation
    • 30+ production solutions delivered across the career; designing, building and maintaining the local data pipelines behind them
    • Anti-hallucination architecture: deterministic R computes every statistic and the LLM only interprets — ~70% fewer tokens, verified by determinism regression tests
    • DevSight: hybrid retrieval over four years of deviation records (2,213 records → 31,577 chunks), ~30 min → ~2 min per search, measured in production; carried into the Horizon proof of concept when DevSight was decommissioned in a global application consolidation
    • BioTrack, the site's manufacturing GenAI platform for microbial investigations: nine-rule deterministic hypothesis engine, ERP maintenance evidence with four kinds of 'nothing', live quality-system feed, 215-assertion regression suite; initial estimate of 450+ hours saved on data aggregation alone
    • Led the AWS S3 → Databricks Unity Catalog migration across two production platforms; authored the URS, the Architecture Council submission and a 16-artifact SDLC dossier

  3. Nov 2022 – Mar 2025

    Data Scientist III (Sr. Manager)

    Takeda Pharmaceutical

    Drove the AGILE 4.0 Digital Workstream and cutting-edge analytics to accelerate Takeda's digital capabilities across GMS sites.

    • Drove the AGILE 4.0 Digital Workstream across all GMS sites
    • Automated JD Edwards commodity deviation monitoring — 59% reduction in deviations
    • Filtration cycle-time investigation: multivariate analytics (Elastic Net, Random Forest, Boruta, SHAP) surfaced the hold-condition drivers a year of univariate work had missed as probable root causes — supported a ~25% cycle-time reduction
    • Executed local and global GMS/GQ big-data and analytics strategy
    • Built and maintained data pipelines and BI applications across all GMS plasma sites

  4. Apr 2021 – Nov 2022

    Data Scientist II (Manager)

    Takeda Pharmaceutical

    Built AI/ML and advanced analytics programs to gain process insight and accelerate digital capabilities.

    • Achieved ~50% reduction in process variation: the ProModel digital twin located the improvement areas and PIMS tracked the gains in near real time
    • TEASEr — TrackWise deviation text analytics and search engine used across GMS sites: 80% improvement in deviation search, closure and analytics; the work behind the 2023 Race4Value award
    • Designed A/B tests and developed end-to-end ML applications
    • Los Angeles lead data scientist for global analytics initiatives

  5. Dec 2019 – Apr 2021

    Data Scientist, Analytics

    Takeda Pharmaceutical

    • Developed and deployed PIMS — real-time analytics across the LA site
    • ML root-cause models cut teardown pooling loss by over 40%, supporting the global immunoglobulin-yield initiative
    • Los Angeles Qlik Sense stream owner and admin since 2017 — eight applications in production, plus the data connections and app-development support other teams build on
    • Statistical advisor; deployed ML models for predictive analytics in R/Shiny

  6. Apr 2018 – Nov 2019

    Operations Leader

    Shire → Takeda

    • Led a $1.2M VIP project reducing equipment waste across the LA site
    • Built BI apps and the PIMS real-time analytics platform
    • Owned the LA Fractionation digital twin (ProModel discrete-event simulation, 2016–2020): micro-step analytics and equations that reproduced floor variation at 95% agreement with production
    • Led DMAIC problem-solving initiatives

  7. Jun 2016 – Mar 2018

    Senior Operations Analyst

    Shire

    • Built PIMS v1 (2016) as a .NET desktop application that recorded manufacturing delay data in one database for analysis
    • Set up the site's process SQL database (2016), linked live to the DeltaV historian replica; still own its connections and stored procedures
    • Qlik Sense stream owner and admin for the Los Angeles site since 2017
    • Took ownership of the LA Fractionation digital twin (ProModel) in 2016; Six Sigma and advanced analytics support

  8. Mar 2015 – May 2016

    Senior Operations Analyst

    Baxalta

    • Transitioned with the Baxalta spin-off from Baxter
    • System development, data analysis and project management

  9. Nov 2013 – Feb 2015

    Operations Analyst

    Baxter International

    • Data analysis and process improvement
    • Early analytics and reporting automation

  10. Sep 2005 – Oct 2013

    Manufacturing Technician 3

    Baxter International

    • 8+ years of hands-on manufacturing-floor experience
    • Foundation in biopharma production processes

Skills

Languages
R, Python, SQL, TypeScript, JavaScript, HTML / CSS, .NET, Java
R Ecosystem
Shiny, bslib / bs4Dash, ellmer, querychat, Quarto / R Markdown, testthat, pool / DBI, reticulate
Data & ML
SPC (XmR, p-charts, Cpk), Elastic Net · Random Forest · XGBoost, SHAP / Boruta, SIMCA / OPLS, Wilson intervals, Base-rate analysis, Multiple-comparison correction, Prophet / changepoint, Apache Spark, Plotly, Dash, Minitab, JMP
Web & Frontend
React, Next.js, Tailwind CSS, Node.js, Vite, Express, Zod, Prisma, shadcn/ui, Auth.js, Supabase (RLS)
Databases
PostgreSQL, SQL Server, Oracle, Supabase, Neon
Cloud & DevOps
AWS, Databricks, Databricks Unity Catalog, Databricks Vector Search, Databricks Apps, Posit Connect, Vercel, Docker, Git, GitHub Actions, Vitest / Playwright
BI Tools
Qlik Sense, Power BI, Tableau
AI & GenAI
Anti-hallucination architecture, LLM Integration, RAG Architecture, Vector Search, Tool Calling, Prompt Engineering, Text-to-SQL, Agentic AI, Multimodal / vision, SSE streaming, Databricks Foundation Models, Claude / GPT APIs, OpenAI / Anthropic / Ollama, Vercel AI SDK, Voice AI (TTS/STT)
Methodologies
Six Sigma, SCRUM / Agile, DMAIC, MLOps, Project Management, SDLC governance (URS, non-GxP informational), WCAG 2.1 AA

Education and certifications

  • 2026Build Data Pipelines with Lakeflow Spark Declarative Pipelines, Databricks Academy
  • 2026Data Warehousing with Databricks, Databricks Academy
  • 2026Building Single-Agent Applications on Databricks, Databricks Academy
  • 2023Oxford Machine Learning Summer School, AI for Global Goals
  • 2023ODSC AI Bootcamp, Open Data Science Conference
  • 2020Data Science Professional, HarvardX (Harvard Online)
  • 2020Statistical Learning, Inference & Modeling, HarvardX / Stanford Online
  • 2012B.S., Computer Science, Cal State Northridge
  • 2004Ingeniero Informático, Informatics, Pontificia Universidad Católica del Perú

Programme roles: Lean Six Sigma Green Belt, AGILE Champion, Databricks Champion, Mulesoft Champion.

Recognition

  • 2023Race4Value — FTE Optimization Winner, GMSGQ award across all plasma manufacturing sites
  • 2022Star of the Quarter, Q3 FY22, Los Angeles site
  • 2017Employee of the Quarter, Los Angeles site

Last updated October 2026. The PDF is generated from the same data.