Mera / NuSeat — AI Job Matching Platform

AI / Job Matching

A job matching platform built at NuSeat using the novel L-Trees algorithm — a dynamically-sorted categorical tree powered by LLMs. The tree's state guides the LLM in asking the right questions to match candidates with job postings. Later evolved into CL-Trees (chronological variant for memory) and ultimately DynamicContextObjects built on pgvector.

Impact & Results

Novel algorithm with broad applicability beyond job matching

Conceptual overlap with Microsoft GraphRAG validates approach

Evolved into reusable DynamicContextObjects library

Challenges & Solutions

Challenge 1

Matching candidates to jobs requires understanding nuanced preferences

Solution 1

L-Trees: dynamic, LLM-powered categorical tree algorithm

Challenge 2

Traditional keyword matching misses semantic similarities

Solution 2

Tree state fed to LLM to guide question selection

Challenge 3

Conversation context needed to guide follow-up questions

Solution 3

CL-Trees for chronological task memory and up-summarization

Technology Stack

Python LLM APIs pgvector Custom L-Trees Algorithm CL-Trees