Experience

AI Product Builder

BCG X

Boston Consulting Group · New York, NY

• Own the quality-evaluation workstream for the CFS Starter-Kit — a reusable-AI-component platform (Project Helix) — building its first scoring rubric and an automated TypeScript/Vitest harness that set the recall / precision / critical-recall baselines the team measures against.
• Build a ground-truth benchmark against the real product modules that resets the platform's headline retrieval metric from 71% on synthetic prompts to ~15% on real builds — exposing a structural coverage cap the synthetic tests had hidden.
• Rebuild block-retrieval scoring from flat keyword overlap to BM25 with field-weighting and stemming, and layer on synonym-mapping and dependency-expansion levers that raise measured recall to ~90% and cut the plain-language vocabulary gap from 38% to 13%.
• Design a persona-by-prompting-skill evaluation that exposes how skill-sensitive retrieval is, then shows the chat's clarifying-question flow recovers weak prompters — lifting novice results from ~35% to 71%.
• Ship block retrieval into the v2 assembly chat and a Claude Desktop MCP server (block search / bundle), plus a relevance floor that returns nothing on off-domain queries instead of a misleading best-of list.
• Surface data-integrity and coverage defects that automated scoring catches but manual review had missed — broken cross-references across 37% of the catalog, and a home-screen module with zero catalogued components.

Independent Contractor — AI & Content

OCI Group Consulting

Remote

• Designed and delivered an 8-module AI enablement program for the firm's partners — deployed as a self-serve web portal (Vercel), not a slide deck — covering Gemini/Claude/ChatGPT workflows, prompt engineering, and AI governance, tailored live to their actual stack.
• Built the credentials deck the firm used in a live pitch for a role on a multi-billion-dollar development project — synthesizing a dozen+ conflicting source proposals into one narrative across 11 client-feedback revisions; the deck advanced the partnership and is being productized into a reusable template.

AI Agent Developer

PwC

Capstone Project × Fordham

• Automated end-to-end M&A deal monitoring across 51 live sources via a RAG pipeline (LangChain, Chroma, GPT-4o) with 8-K parsing, MD5 dedup, and daily scheduling.
• Built a structured extraction and SEC XBRL enrichment pipeline, cutting deal-attribute extraction from hours to sub-minute output while standardizing 10+ fields.
• Delivered a self-service dashboard replacing ad-hoc research requests with a persistent intelligence layer.

AI Graduate Consultant

United Nations

Office of Legal Affairs · with Fordham

• Delivered UN-LEX as a production tool adopted by the UN Office of Legal Affairs as its primary research interface across 9,000+ UNDT/UNAT cases spanning 15+ years of precedent.
• Reduced case-law research time from hours to minutes — on-demand precedent lookup by fact pattern or misconduct type, with paragraph-level citations, delivered inside Microsoft Teams.
• Captured senior officers' precedent expertise into a durable, queryable system — cutting onboarding time and mitigating operational risk from attorney turnover.

Internship · Tutoring · Public sector

Earlier roles

• Lead Tutor, First Generation Investors — led a team of four tutors delivering weekly personal-finance sessions (stocks, bonds, diversification, dollar-cost averaging) at two San Antonio high schools; guided students through a capstone where each invested $100 of real money.
• Municipal Intern, Bexar County — documented weekly court proceedings under Judge Melissa Saenz (arraignments, bench trials, sentencing), researched local/state/federal statutes, and discussed case law and judicial reasoning directly with the judge.
• Procurement Analyst Intern, Security Service FCU — managed contracts, POs, assets, and vendor records in System Center Service Manager, and ran IT device lifecycle: imaging, deployment, and secure decommissioning of PCs, laptops, and phones.