Skip to main content
Back to HomeAdam Silva

Featured Project

Upskill

AI-powered learning platform that bridges daily work output and professional development — detecting skill gaps automatically, delivering learning exactly when it matters.

Upskill dashboard overview
01 / 05

Overview

Upskill is an enterprise-grade SaaS platform designed to bridge the gap between daily work output and professional development. By integrating directly into a team's digital ecosystem via Slack and Google Workspace, the platform autonomously detects skill strengths and gaps — without any manual input.

The core intelligence is a dual-model pipeline: Llama 4 handles deep analysis of work artifacts across its expanded 10M token context window, while GPT-5 synthesises those insights into personalised, actionable learning interventions. The entire stack is deployed on Microsoft Azure Foundry, providing enterprise security, PII protection, and global low-latency access.

Tech Stack

Analysis Engine

Llama 4

Content Generation

GPT-5

Cloud Infrastructure

Microsoft Azure Foundry

Frontend

Next.jsReactTypeScriptTailwind CSS

Backend

PythonFastAPISQLAlchemyPydantic

Database

PostgreSQLAlembic

Key Architectural Features

  1. 01.

    Dual-Model Intelligence Pipeline

    • Llama 4 (MoE) processes massive volumes of daily work artifacts — messages, documents, code snippets — using its 10M token context window for long-term pattern recognition.
    • GPT-5 transforms raw analytical data into nuanced, highly contextual learning interventions tailored to each individual's specific project history and skill trajectory.
  2. 02.

    Enterprise Deployment on Azure Foundry

    • Seamless model orchestration routes requests between Llama 4 and GPT-5 based on task type — analysis vs. content generation.
    • Enterprise-grade content filtering and PII protection applied to all ingested work artifacts before analysis.
    • Azure's global footprint ensures low-latency real-time analysis of communication data at scale.
  3. 03.

    Modern Full-Stack Implementation

    • High-concurrency FastAPI backend with a dedicated service layer for artifact ingestion, asynchronous analysis pipelines, and strict Pydantic validation.
    • Responsive Next.js dashboard with a custom design system supporting high-contrast accessibility and a hook-based data layer for real-time Skill Intelligence updates.
  4. 04.

    Automated Proof of Work Mapping

    • AI detects objective skill mastery from daily artifacts — no self-reported surveys or manual input required.
    • Creates a dynamic Taxonomy of Strength providing leadership with a data-driven view of organisational capabilities.

Key Impacts

Zero-Friction Ingestion

Automated connectors for Slack and Google Workspace eliminate manual data entry, making skill tracking invisible to the end user.

High-Fidelity Diagnostics

Llama 4 identifies subtle skill gaps and emerging strengths that traditional analytics tools miss, using long-horizon context across weeks of work artifacts.

Just-in-Time Learning

GPT-5 generates targeted interventions delivered exactly when a skill gap is detected, maximising professional development ROI without disrupting workflows.

Back to Home

Designed & Built by Adam Silva