Noelle Milton Vega - Resume

Consultancy focused on enabling business-critical technology platforms and architectures with Agentic Generative AI. All engagements are performed via Jupyter, LLC unless noted otherwise.

Automated document compliance analysis with LangGraph, LangChain.

For a Fluree financial client, I developed a LangGraph-powered document compliance analysis agent that analyzed thousands of in-force vendor and client communications - including flagging deprecated terms, signatures from departed executives, and even outdated logos. The solution leveraged LangChain for supplemental capabilities (e.g., chat templates, ChatOllama, etc.), Postgres with pgvector to store relational data as well as to semantically index reference documents against which target documents were compared. Pydantic models mirroring the Postgres schema were programmatically generated (via a db introspection script) and injected into the LLM system prompt, enabling it to generate SQL queries from natural language user queries. This solution replaced manual offshore third-party review teams, cutting project costs by 63% while eliminating human error and significantly decreasing time-to-completion. It also established a reliable audit trail and a reusable system for future operations.

Spearheaded and delivered project GOWL (GPU + Ollama + OpenWeb UI + LLMs) for a financial services client.

For a financial services client, I designed and delivered Project GOWL - a laptop replacement initiative equipping every employee with AI-enabled devices featuring GPU acceleration, Ollama for serving local LLMs, Open-WebUI for chat interfaces, and select LLM models. This end-to-end solution elevated corporate communications by enabling AI-assisted content creation for Outlook emails, PowerPoint presentations, SharePoint/Word documents, and Teams messages - producing more concise yet effective content while operating entirely within the secure, compliant boundaries of a local LLM architecture. This avoided the security, risks, and regulatory exposure associated with cloud-based chat solutions.

Led a dual-phase digital transformation project.

Led a dual-phase digital transformation initiative that modernized an enterprise-critical function from error-prone spreadsheets to Adobe Workfront’s collaborative platform, enabling end-to-end project management, workflows, tasks, approvals, messaging, custom forms, and lineage tracking. For project Phase 1 (Assessment), I analyzed business requirements and dataflows, authored user stories for POCs, and selected Workfront based on technical feasibility, functional fit, cost efficiency, time-to-value, usability, and low-code flexibility. For project Phase 2 (Implementation), I worked with Adobe to deploy Workfront, configured business user stories, documented processes, and trained users—successfully transitioning the mission-critical workflow to a scalable, governed platform.

AWS platform architect and developer for the ZettaSense cloud product suite.

At ZettaLabs (now FLUR.EE), I contributed to the flagship ZettaSense product suite - including ZettaSense Ingest (data streaming logistics), ZettaSense Classify (discovery and classification of raw data into business domains), and ZettaSense Resolve (mastering, merging, and deduplication). For this small, niche company serving Financial Services and Pharmaceutical clients, I wore these hats: (1) Designing cloud data architectures and platforms for product deployments, (2) Delivering professional integration services for clients/adopters, and (3) Developing PySpark and Kafka solutions for ZettaSense components.

Directed teams in US and Canada focused on Data Lake platform for Office of Data Governance transformation initiatives.

Reporting directly to the global Chief Data Officer within the Data Governance Office, I led US and Canadian teams focused on a Cloudera Data Lake platform supporting enterprise data governance initiatives. This platform served as the single source of truth for Customer and Transaction data across all bank lines of business (LOBs), where LOB data was coalesced, mastered, merged, and quality-monitored through user stories. Key technologies included: Attunity/Qlik Replicate & Compose for database CDC; Sqoop and Apache Kafka for data logistics and lake ingest (including CDC streams); PySpark/Python with libraries for batch/streaming analytics atop Spark 2.x; HiveQL and SparkSQL for querying; and Jupyter Notebook for analysis. The platform's programmatic interaction model aligned with my published tutorial, developed using a personal Data Architecture & Analytics R&D Lab.

Tutorial: https://jupyter.ai/data-pipeline
R&D Lab: https://jupyter.ai/rndlab

Designed and built Complex Event Processing solution to replace aging FIX trade analytics system.

Engaged by Bank of America's trading division to assess requirements for replacing their legacy FIX trade analytics system, and to propose, secured buy-in for, and build a Complex Event Processing solution. Using open-source technologies including, Apache Storm, Cassandra, and Redis, I designed and delivered a streaming platform that would be used to identify wash sales, orders-to-execution latencies, and account limit breaches in real-time. It was capable of processing 12-million events per second, demonstrated the viability of a "Open Source Build" over "Buy" approach bank-wide. Here's a post detailing some challenges I encountered along with solutions for them:

Apache Storm Continuity-of-Service: https://jupyter.ai/bofa

Performed the design and implementation of a comprehensive Business Continuity and Disaster Recovery solution.

Engaged by NYMEX to design and implement a post-9/11 Business Continuity & Disaster Recovery (BC/DR) platform, I delivered a solution featuring a WAN/SAN connecting primary and backup datacenters for real-time synchronous replication of commodities trades and critical database transactions - mitigating operational, financial, and legal risks through measurable KPIs (RPO/RTO). These were validated in bi-annual failover tests, which I subsequently authored. My full-cycle leadership included: Presenting solutions to C-level executives (CEO/CIO/CTO) to gain consensus; Justifying CAPEX/OPEX via ROI/TCO analysis to secure 3-year budget approval; Designing pay-as-you-grow infrastructure to spread CAPEX; Vendor evaluation; Collaborating with facilities teams and trade unions; Architecting the inter-datacenter Storage Area Network platform; And documenting all policies/procedures. These efforts enabled NYMEX’s disaster resilience, as reflected in the following platform design documentation:

Cross Datacenter SAN BC/DR platform design: https://jupyter.ai/nymex