💬
Evie Avatar Ask Evie
Evie: Hi, I'm Evie 👋 — your digital assistant from Evangelist Apps. Ask me anything about our services, projects, or tech stack! If you'd like us to get in touch, share your contact details below.

Experience Evangelist’s value across end-to-end digital deliveries for key industries.

Contact Us
Contact Us

CASE STUDY

Transforming Expert Discovery with AI-Powered Semantic Search 

  • Client: Internal Product – Evangelist Apps 
  • Industries Focused: Recruitment / Talent Intelligence 
  • Solution: AI-Based Expert Identification
  • Technology Stack: React, NextJS, Django, PostgreSQL, FAISS, Sentence Transformers, MiniLM, Python, LLM (GPT/Mistral) 

The Challenge

Traditional expert discovery systems often depend on rigid keyword filters, static taxonomies, or manual Boolean searching. These approaches face major limitations:

  • Limited discoverability of domain-specific experts
  • Inconsistent results from human-led queries
  • Poor scalability with growing datasets
  • Lack of semantic understanding across industries

Evangelist Apps recognized an opportunity to create disruption by building a zero-friction, AI-powered search engine that balances both human intuition and deep semantic intelligence.

The Solution: Expert Finder – A Next-Gen Semantic Match Engine 

We developed Expert Finder as a fully private, on-prem or hybrid AI-powered platform designed for high-speed discovery, semantic matching, and contextual analysis. It goes far beyond keyword search by leveraging advanced language models, embeddings, and real-time relevance scoring with human-in-the-loop feedback. 

Key Innovations 

  • 1. Semantic Natural Language Search
  • AI parses unstructured queries and matches experts using embeddings, rather than exact keyword matches.
  • 2. LLM-Powered Prompt Expansion
  • Using custom-tuned LLMs (e.g., GPT-4 or Mistral) to expand user prompts into richer content trees – including synonyms, adjacent skills, and industry-specific jargon. 
  • 3. Context-Aware Hybrid Backend Architecture 
  • Bi-Encoder (FAISS) + Cross Encoder (MiniLM or MPNet) to score how well each expert aligns with a query. 
  • 4. LLM-Generated Insights for Matched Profiles
  • Each search result carries a match explanation summary, skill-based score (0–100), and an overall fit score for ranking transparency.

Features & Capabilities

  • 1Skill-Level Scoring
    Each expert match is scored at skill, domain, and overall fit level. 
  • 2Expansion Matrix
    Query expansion into adjacent areas with clear visual indicators.
  • 3Query Refinement
    Human-in-the-loop to further fine-tune AI recommendations. 
  • 4Sub-Second Latency
    AI index optimized for <1s response time, even on 10K+ expert corpora.

Technical Architecture

A future-ready modular stack designed for scale, privacy, and integration:

  • Frontend: React, Next.js, Tailwind CSS
  • Backend: Django + Django REST Framework (DRF)
  • Database: PostgreSQL (main), SQLite (dev/light use)
  • AI/ML Stack:
    • FAISS – Fast vector similarity search
    • Sentence Transformers (MiniLM, MPNet) – for embeddings
    • LLMs (GPT-4/Mistral) – for query expansion and match summaries
  • Deployment: Dockerized services (optionally cloud-deployable) 
  • Monitoring/CI: GitHub Actions for build + test automation 

Evangelist Apps recognized an opportunity to disrupt this paradigm by building a zero-friction, AI-powered search engine that mirrors how humans think — not how databases filter. 

Results

The Expert Finder platform is currently in internal deployment, with highly encouraging benchmarks:

  • 40% faster expert identification than traditional filters 
  • 55% higher match accuracy (based on internal QA scoring)
  • 90% integration done with any normalized expert database schema

What’s Next

  • 1. Client Rollouts:
  • Pilots underway with expert networks and recruitment agencies 
  • 2. Multilingual Search:
  • Expanding LLM capabilities to non-English queries
  • 3. Knowledge Graph Overlay 
  • To enrich link clustering and profile relationships
  • 4. Analytics & Dashboards
  • Upgrading metrics for expert pipeline attrition and usage
  • 5. On-Device AI & Cost Optimization
  • Exploring quantized LLMs for lower compute load

Conclusion 

Expert Finder by Evangelist Apps is more than just a search tool – it’s a semantic match engine that blends structured and unstructured data into a real-time expert discovery experience. Whether used in recruitment, expert networks, or specialist advisory domains, this platform offers speed, precision, and unmatched explainability. We are continuously improving the stack and roadmap for external deployments in 2025. 

  • CONTACT
    +44 (0) 2080951050

London Office

Cressex Business Park
Lincoln Road,
High Wycombe,
HP12 3RL

Fleet Office

Sentinel House
Ancells Business Park,
Harvest Crescent,
Fleet GU51 2UZ

Services

Technologies

Resources

Industries

  • Education
  • SaaS
  • Finance
  • On-Demand
  • eCommerce
  • Logistics

Portfolio

Full stack mobile (iOS, Android) and web app design and development agency

Terms & Conditions

Medium X-twitter Linkedin