Li MingSenior Software Engineer
96% Back to homeTalentAI / Talent Operating Console
Turn resume libraries into explainable talent discovery systems
TalentAI is not a search box with an LLM. It is a hiring intelligence workspace for resume parsing, hybrid recall, JD scoring, evidence explanation, and private deployment.
Role brief
Head of AI Engineering
Needs LLM application experience, vector retrieval, cloud-native deployment, and cross-functional delivery.
Ranked candidatesEvidence-driven
Wang JingProduct Manager
92%Chen LeiML Engineer
89%1,245Candidate profiles
< 3sJD match response
RRFFusion ranking
PrivateDeployable privately
Product position
From finding resumes to explaining why they fit
The hard part is not listing candidates. It is helping hiring owners, HR, and delivery teams understand why a person deserves the next conversation.
Resume parsing is not OCR
PDFs become computable talent profiles with experience, skills, projects, and timelines.
Keyword and semantic search work together
Exact terms, Chinese tokenization, vector recall, and RRF fusion decide ranking together.
Matches are explainable
Scores come from role requirements, evidence, ability tags, and risk signals.
Designed for private data
Containerized deployment keeps resumes, JDs, and match history in controlled environments.
Workflow
One full path from resume to candidate explanation
Resume intake
PDF, Word, and text resumes enter one parsing queue with raw and structured records.
LLM parsing
Experience, skills, companies, tenure, education, and project context become searchable profiles.
Hybrid recall
BM25, zhparser, and pgvector run together for keyword precision and semantic recall.
JD scoring
The LLM evaluates fit by dimension and returns scores, rationale, and risk signals.
Explainable handoff
Evidence snippets and next actions become decision material for hiring teams.
Architecture
Hybrid search, vector recall, and LLM explanation are governed separately
This is not one prompt doing everything. TalentAI separates structured data, Chinese full-text search, vector indexing, and model scoring into replaceable layers.
Interface proof
This is not a deck. The core surfaces are demoable.
Screenshots are evidence of capability. The website explains the product logic, while the subdomain carries the live walkthrough.
Talent pool view
Scan candidates by skill, company, tenure, and match quality.
JD match view
Role requirements and candidate scores live in the same decision surface.
Resume parsing view
Uploads become structured experience and skill tags with less manual entry.
Delivery
Ready for online trial and private deployment
Suitable for search firms, internal hiring teams, talent mobility, and executive search. Public pages explain capability; customer data stays in controlled environments.