AI Knowledge Assistant & LMS Platform

preview

AI-Powered Knowledge Platform and Training for Distributed Teams

About project

An internal knowledge base, AI assistant, and microlearning LMS built for a mid-sized logistics company to streamline employee onboarding, self-service support, and knowledge access.

Challenge

Our client, a logistics company with over 150 employees across 7 locations in the U.S., faced increasing inefficiencies in onboarding, support, and internal communication.

Key challenges included:

  • Scattered Information: Company policies and procedures were split between Google Docs, email threads, and team memory.
  • Inconsistent Onboarding: New hires lacked centralized training, causing delays and operational risk.
  • Support Overload: HR, IT, and team leads were bogged down by repetitive questions about tools, processes, and policies.

The company needed a unified system to provide employees with self-service access to verified knowledge and flexible training across office and field teams.

Solution

1. Centralized Knowledge Base

We built an internal portal to consolidate SOPs, policies, FAQs, and guides — categorized by department, access-controlled by role, and fully searchable via semantic indexing.

case-photo

2. LLM-Powered Assistant

We deployed a retrieval-augmented assistant trained on internal documents using OpenAI embeddings and a Qdrant vector database. Instead of full fine-tuning, we used dynamic retrieval and context-based prompting for scalability and faster rollout.

The assistant was accessible via:

  • Web Chat Widget — embedded in the knowledge portal
  • Slack Bot — integrated directly into the company’s Slack workspace for instant access within channels

Permission-based access controls ensured that employees only received responses relevant to their role and location.

case-photocase-photo

3. Lightweight Custom LMS

To accelerate training, we developed a mobile-friendly LMS with short modules on:

  • Delivery reporting
  • Dispatch protocols
  • Time-off procedures
  • Safety and compliance workflows

Each course featured short videos, step-by-step visuals, and embedded quizzes.

case-photocase-photocase-photo

4. Why a Custom LMS?

Existing LMS tools (e.g., LearnUpon, TalentLMS) were evaluated but lacked:

  • Seamless integration with the AI assistant and knowledge portal
  • Mobile-first UX for field workers
  • Real-time control over content updates and logic
  • Flexible analytics tied to internal processes

Our custom solution allowed fast content iteration, embedded assistant guidance, and full control over UX and data privacy.

5. Integrations

The platform connected with Slack, Google Workspace, Notion, and Keycloak (SSO). All assistant queries and LMS progress were tracked in a unified analytics layer for HR visibility.

case-photo

Project Team

  • 1 Project Manager
  • 1 AI/NLP Engineer
  • 2 Backend Developers
  • 1 Frontend Developer
  • 1 UX/UI Designer
  • 1 QA Engineer
  • 1 DevOps Engineer

Tech stack

openai
qdrant
react
tailwind
fast-api
postgresql
nodejs
slack
notion
google-workspace
keycloak
docker
aws
kubernetes
github-actions

Timeline

Total duration: ~14 weeks (MVP scope)

Weeks 1–2

Discovery, stakeholder interviews, content audit

Weeks 3–4

UX flows, KB structure, semantic indexing setup

Weeks 5–6

Embedding pipeline, Slack/web assistant integration

Weeks 7-8

RAG testing, permissions logic, assistant validation

Weeks 9-10

LMS development, content upload, mobile optimization

Weeks 11-12

Final deployment, pre-event support, on-site readiness

Weeks 13-14

Optimization, onboarding support, internal training