Back to projects
AI AgentsSemantic SearchData Engineering

PwC: CV Semantic Search Platform

An internal talent-search platform using semantic retrieval to make staffing and proposal work faster.

Role
Search and internal tooling engineering
Focus
AI Agents / Semantic Search
Stack
Python, FastAPI, Docker

Context

The shape of the problem

Finding the right internal expertise is a search problem with messy inputs: CV formats vary, terminology differs by team, and managers need useful matches quickly. The goal was to turn that unstructured material into a search experience that reflected actual skills and project history.

Build

What I actually built

I worked on the ETL and semantic-search pipeline, standardizing CV data, embedding relevant content, and building an interface for practical discovery. The system focused on retrieval quality, speed, and adoption by non-technical users.

Engineering

Key decisions

01

Automating extraction, transformation, and loading of diverse internal resume formats

Developed an automated ETL pipeline using Python and Docker to fetch and standardize resume data

02

Implementing a robust semantic search system capable of accurately interpreting skills and experience

Implemented a sophisticated RAG pipeline and semantic search for accurate talent matching

03

Creating an intuitive and accessible frontend for non-technical consulting managers

Built a user-friendly Streamlit interface enabling easy and efficient talent discovery

04

Ensuring data privacy, compliance, and secure access management

Established secure role-based access controls with detailed audit logging

Outcome

What changed

  1. 01Significantly accelerated talent identification processes for consulting managers
  2. 02Improved accuracy and efficiency in staffing projects based on internal skills
  3. 03Enhanced internal resource utilization and mobility across projects
  4. 04Successfully deployed and adopted across multiple departments

Stack

Tools and systems

PythonFastAPIDockerStreamlitRAGSemantic Search

I build delightful systems.

© 2026 Kristian Ernst.