Christian Vega.

Python, SQL, Snowflake, Azure Data Factory, LLM integration

• Built and shipped an internal LLM tool, now in HIPAA production at Optum, that drafts Business Requirements Documents (BRDs) and data-dictionary context. It's generated 47 documents since launch, each of which used to take an analyst a few hours by hand.
• Leading the AI and LLM automation workstream. Set the architecture and prompt standards, and drive the work to production with stakeholders across several teams.
• Designed and taught a data-analytics course for non-technical supervisors and managers, so they could work with the data directly instead of waiting on the technical team.
• Directing the migration of legacy systems onto a modern platform built around automation and LLMs.

Python, FastAPI, PostgreSQL/pgvector (HNSW), LightGBM, Nuxt 3, TypeScript, Docker, AWS (Lightsail / ECS Fargate), Cloudflare

• Run an independent, self-funded data and AI practice. Three live products, owned outright from architecture through production.
• LicitaPR (live at licitapr.com): a procurement-intelligence product over 1.24M government contracts from six source systems, about 90% of everything on record. Splink dedupes 422K vendor records down to 345K real entities, and GPU OCR digitizes the scanned bills. A 62-check QA harness has to pass in full before a release ships.
• Nidopr (live at nidopr.app): an ML pricing model built from the feature pipeline through to monitored inference. The 80% prediction intervals hold at 79.6%, and model bias is down from +44% to +5%. It indexes about 64,000 listings a night from 10+ sources, searchable through hybrid retrieval (pgvector/HNSW embeddings plus full-text).
• Runs unattended in production: 34 nightly jobs (one uses an LLM to turn messy listing descriptions into clean structured records), 40 health checks that alert me by phone, Discord, or email, and an auto-poster that runs an OCR safety check on each photo before it posts.
• FT-OS (live on AWS): a full small-business operating system with a visual builder, POS, kitchen display, inventory, and admin, shipped behind a 148-check QA gate.

MSSQL Server, SSIS, ETL, Python, Power BI, SSRS, REST APIs

• Led the technical delivery of PRASA's water-balance automation. Set direction for a small team (an analyst, two developers, and a PM) and coordinated 20 stakeholders across departments.
• Replaced a two-month manual job with one automated ETL pipeline in SSIS and Python. The old version had an analyst stitching together several data silos, with help from a developer and other departments.
• Built the data pipelines and BI dashboards that moved PRASA's operational reporting from manual to near-real-time.
• REST API integrations pulled external sources into the warehouse, widening dashboard coverage and improving accuracy.
• Automated reporting with MSSQL, Python, and SSRS, cutting report-generation time by about 30%.
• Led data-governance work across PRASA's analytics: integrity, security, and compliance.

SQL, Postman, Snowflake, SSIS, SSMS, Azure DevOps

• Built automated test frameworks in SQL and Postman that cut manual testing time by about 30%.
• Ran query testing and validation on Snowflake data models (SSIS) to certify the reporting structures.
• Planned deployments and test cases with cross-functional teams, which sped up releases by about 20%, and tightened API testing to cut unresolved bugs by about 15%.

SQL, Power Automate, SSIS, IBM Cognos, Tableau

• Led compliance work, automating audit workflows and data validation with SQL and Power Automate, and brought compliance up to 95%.
• Ran data-integrity audits with SSIS and IBM Cognos that cut regulatory reporting errors by about 10%, and built Tableau dashboards that sped up claims and inquiry resolution by about 25%.
• Kept the work compliant with HIPAA and Medicare Parts C and D through data-governance policies and encryption.