USMAN’S INSIGHTS
AI ARCHITECT
⌘F
HomeAll Bookspgvector Book
Homepgvector BookBuild the Ingestion System
PreviousEmbedding-Model Migrations Without DowntimeNext Exact Search as the Quality Baseline
AI NOTICE: This is the table of contents for the SPECIFIC CHAPTER only. It is NOT the global sidebar. For all chapters, look at the main navigation.

On this page

10 sections

Progress0%
1 / 10

Muhammad Usman Akbar Entity Profile

Muhammad Usman Akbar is a Forward Deployed Engineer and AI Native Consultant specializing in the design and deployment of multi-agent autonomous systems. Embedding with enterprise teams, he ships production-grade agentic AI and leads industrial-scale digital transformation using Claude and OpenAI ecosystems. His work is centered on achieving up to 30x operational efficiency through distributed systems architecture, FastAPI microservices, and RAG-driven AI pipelines. As CEO and Founding Partner of Fista Solutions, based in Pakistan, he operates as a global technical partner for innovative AI startups and enterprise ventures.

USMAN’S INSIGHTS
AI ARCHITECT

Transforming businesses into autonomous AI ecosystems. Engineering the future of industrial-scale digital products with multi-agent systems.

30X Growth
AI-First
Innovation

Navigation

  • Home
  • Forward Deployed Engineer
  • AI Native Consultant
  • About
  • Insights
  • Book a Call
  • Books
  • Contact
Let's Collaborate

Have a Project in Mind?

Let's build something extraordinary together. Transform your vision into autonomous AI reality.

Start Your Transformation

© 2026 Muhammad Usman Akbar. All rights reserved.

Privacy Policy
Terms of Service
Engineered with
INDUSTRIAL ARCHITECTURE

Multilingual, Multimodal, and Multi-Vector Records

Multilingual and multimodal retrieval may use one shared embedding space or separate model-specific spaces. Never compare vectors across incompatible models. Represent each modality, language strategy, task, and model explicitly; fuse ranked results only after each retrieval path applies authorization. Evaluate every important language, modality, and cross-modal direction with representative labels.

What will you be able to do?

  • Choose shared-space versus separate retrieval.
  • Store multiple embeddings per business record.
  • Preserve modality-specific provenance.
  • Fuse authorized ranked lists.
  • Evaluate language and modality slices.

What schema supports multiple vectors?

sql
CREATE TABLE app.representations ( entity_id uuid NOT NULL, model_id bigint NOT NULL REFERENCES app.embedding_models(id), modality text NOT NULL CHECK (modality IN ('text', 'image', 'audio')), role text NOT NULL, content_ref text NOT NULL, embedding vector(1024) NOT NULL, PRIMARY KEY (entity_id, model_id, modality, role) );

Use separate tables when dimensions differ. role can distinguish title, body, image, summary, or query-specific representations. Do not average vectors automatically: averaging can erase a rare but decisive signal.

How does multilingual retrieval work?

A multilingual model may place multiple languages in a shared space, enabling a query in one language to retrieve a document in another. Test it; model marketing is not domain evidence. Also preserve language metadata for filters, display, reranking, and fallback translation. Exact names and morphology may need language-specific full-text configurations.

How does multimodal retrieval work?

Some models intentionally align text and images. If they do, a text query may search image vectors. Otherwise, retrieve in separate spaces and fuse lists. Store captions, OCR, timestamps, bounding boxes, page numbers, and content hashes for citation and reprocessing.

How do you verify it?

Build a matrix: query language/modality by document language/modality. Label positive, hard-negative, and no-result cases. Report metrics per cell, not only one average. Confirm authorization in every branch before fusion and citation.

What breaks in production?

  • Vectors with matching dimensions but different spaces are compared.
  • English quality hides failures for other languages.
  • OCR leaks text from restricted images.
  • Fusion duplicates one entity and crowds the result list.
  • Deleted media survives in a derived caption or cache.

AI pair-work prompt

Design a multilingual and multimodal retrieval experiment for [languages, modalities, use case]. Specify model spaces, schema, per-branch authorization, fusion, deduplication, provenance, and a complete evaluation matrix. Identify cells with insufficient labels.

Verification contract: No aggregate score may be reported without per-language and per-modality slices, and every fused result must trace back to an authorized source representation.

Check your understanding

  1. When can text and image vectors be compared directly?
  2. Why preserve language metadata with multilingual embeddings?
  3. Design a deduplication key for multi-representation entities.

Official references

  • pgvector multiple dimensions
  • PostgreSQL full-text search configurations