Summary

AI technical lead who designs and ships LLM applications and production AI agents, and owns whether the work actually landed. I build agentic systems with tool calls, per turn context assembly, persistent memory and long running sessions, and runtime redirection by non engineers without code changes. I design evaluation systems and keep the harness trustworthy enough for the metrics to mean something, telling a broken eval harness apart from an underperforming model. I work AI assisted (build with Claude Code daily, with informed views on its strengths and limits), read code (Python, TypeScript), and make product, UX and build vs buy calls on problems that arrive without a spec. I run on Azure with Docker and CI/CD. 15+ years shipping software that real users rely on. English C1.

Experience

Head of AI · Production AI Agents, Evals & AI Assisted Delivery · DigitalHubAssist LLC Oct 2024 · Present

Remote · Applied AI / SaaS (Anthropic partner program)

Built and own Agent Squad, production AI agents with tool calls that write to the database, per turn context assembly, persistent memory and long running sessions, steerable at runtime by non engineers without code changes (system prompts, tool schemas, guardrails). I build retrieval (embeddings, pgvector) and apply advanced prompt engineering. Under my leadership the company reached Registered Partner status in Anthropic's partner program.
Design and expand evaluation systems that track whether agents improve or regress, keeping the harness trustworthy (LLM judges, deterministic checks, threshold calibration) with observability (Langfuse); I tell a broken harness from an underperforming model. I work AI assisted (Claude Code), read code (Python, TypeScript), run on Azure with Docker, CI/CD and event driven orchestration (Inngest), and integrate systems through MCP tools and function calling.

Consulting Manager · Product & Architecture Judgment, Delivery at Scale · NTT DATA Colombia Apr 2021 · Oct 2024

Bogotá, Colombia · Global systems integrator (GSI) · Financial services (banking)

Owned architecture direction and build vs buy calls on the office's largest account (up to 50 people): partnered with leadership on technical risk, made product calls without a spec, and shipped a platform to production on Azure integrating internal and external systems, with clear post launch outcomes. Recovered a program in crisis (recognized at the corporate level); lowest attrition on the account.

Project Manager / Delivery Lead · Products Real Users Rely On · VASS (2018·21) · GlobalHitss (2015·18) · Tecnocom (2012·15) 2012 · 2021

Bogotá, Colombia · Digital delivery

Shipped products real users rely on: a pension fund's channels (millions of users, LATAMDIGITAL 2019 award), digital commerce and public sector platforms, from design to launch, and stayed connected to whether the work landed. Recognized Liferay expert.

Core Competencies

LLM Applications & Production AI Agents (Tool Calls · Context · Memory · Long Sessions) Runtime Redirection by Non Engineers · Agent Behavior (System Prompts · Tool Schemas · Guardrails) Evaluation Systems (LLM Judges · Deterministic Checks · Threshold Calibration) · Harness Trust Retrieval (Embeddings · pgvector) · Prompt Engineering · Observability (Langfuse) AI Assisted Development (Claude Code) · Reads Code (Python · TypeScript) · Git · CI/CD Azure · Docker · Event Driven Orchestration (Inngest) · MCP Tools · Function Calling Product, UX & Build vs Buy Calls Without a Spec · Technical Risk · Architecture Direction Ships Products Users Rely On · Systems Engineer (STEM) · MBA · English C1

Summary

Experience

Education

Core Competencies