Practicum · GoTyme Bank

My AI Engineering Practicum

A 480-hour AI Engineering practicum inside GoTyme Bank's AI & Automation Pod—shipping project intelligence, knowledge systems, and production-ready AI workflows across a distributed digital bank.

AI Engineer Rookie · January–July 2026

Engineering the systems around the model.

I worked across agentic orchestration, enterprise data, full-stack delivery, ontology design, and handover—collaborating with teams in the Philippines, Vietnam, and South Africa.

AI & Automation Pod

Wince Larcen Rivano at the GoTyme Bank office

480

practicum hours

January–July 2026

global regions

Philippines · Vietnam · South Africa

01 · Primary output

Project intelligence
for product leadership

TymeKeeper turns scattered signals into decisions.

Product Owners needed one reliable view of updates, blockers, sprint progress, team capacity, and decisions—without manually reconstructing context from Slack, Jira, and Confluence.

TymeKeeper project-intelligence dashboard showing alerts and weekly focus

The decision interface

A grounded view of urgent alerts, priorities, project updates, and team context.

Production adoption

TymeKeeper moved from solution design to a deployed Databricks App used by both AI Product Owners.

Incremental memory

Canvas hash-diffing avoided unnecessary LLM calls while preserving project context across refreshes.

Multi-source grounding

Slack, Jira, and Confluence were normalized into one view of updates, blockers, and decisions.

Maintainable handover

Architecture, workflows, tables, deployment, and operating guidance were documented for continuity.

02 · System architecture

n8n · Databricks · Claude
React · Slack · Confluence

A production pipeline, not a prompt wrapper.

Reusable configuration, asynchronous jobs, persistent project memory, and scheduled briefings let one architecture serve multiple project portfolios without rebuilding each workflow.

System architecture and data flow

From enterprise sources and n8n orchestration to Databricks persistence, LLM synthesis, and the frontend.

Zoomed-out view of TymeKeeper's production n8n workflows

Production workflow surface

The orchestrated workflow family behind ingestion, memory refreshes, dashboards, and briefings.

03 · Working in GOOS

Beyond TymeKeeper

Building the knowledge layer around AI.

The practicum also extended into ontology discovery and meeting capture—making organizational knowledge easier for AI systems and people to retrieve, structure, and reuse.

GOOS ontology and schema design

Translated pod context into reusable entities, relationships, and mapping rules for a company-wide knowledge graph.

Meeting Knowledge Capture retrieving a Teams transcript and preparing a Confluence recap

Capture and structure

A purpose-built skill retrieves a Teams transcript and turns it into a reusable meeting artifact.

Meeting Knowledge Capture output with a published recap and locally saved transcript

Publish and preserve

The structured recap is published to Confluence while source artifacts are retained for traceability.

04 · 480 hours

Formal practicum allocation

Time spent where production work compounds.

Delivery led the engagement, supported by documentation, knowledge systems, research, tooling, and structured onboarding.

AI Engineering Work

246h

Technical documentation

72h

Self-Training, Upskilling & Research

72h

GOOS & Forward-Deployed Engineering

50h

Environment and tooling

24h

Orientation and discovery

16h

What I Learned

Production AI is the engineering around the model.

Useful AI systems depend on reliable pipelines, reusable configuration, efficient memory, failure handling, grounded sources, usable interfaces, and clear handover. Technical sophistication matters; adoption is what proves the system creates value.

Wince Larcen M. Rivano · AI Engineer Rookie

WINCE LARCEN