Daem Azeem

AI Solutions Architect specializing in Agentic Systems and Multi-Tenant LLM Architecture.

Based in: Lahore, PKCurrently: Senior AI Developer, Vyera AIExperience: 2 years

Hire me

An experienced AI Solutions Architect and Engineer who designs and ships enterprise-grade agentic systems, specializing in scalable multi-tenant architectures and knowledge graph-driven memory. He excels at translating complex business requirements into robust production-ready AI pipelines using frontier models and retrieval-grounded context.

Experience

Senior AI Developer

Vyera AILahore, PKworkAug 2026 – Present

Summary

Vyera - AI Co-Founder Platform (24+ Autonomous Agents) Python • Agentic AI • Multi-Agent Orchestration • Web Crawling • LLM Orchestration (Claude, GPT, Gemini) • Tool Use • Databases

What I did

  • Own the AI core of Vyera, an AI co-founder platform that runs 24+ specialized agents covering competitive intelligence, SEO/GEO/AEO visibility, marketing strategy, and content execution for SMB and B2B customers.
  • Design and build tool-using agents that autonomously discover and track competitors, crawl and monitor competitor websites, and extract structured metadata (pricing, positioning, product changes) into a persistent database powering real-time alerts and battle cards.
  • Architect the agentic pipeline that turns raw market and competitor signals into brand voice generation, marketing plans, and go-to-market content-routing each task to the frontier model best suited to it (reasoning, writing, speed, or execution).
  • Build the intelligence layer behind AI-visibility tracking (citation and answer-engine monitoring across ChatGPT, Claude, Gemini, Perplexity, and AI Overviews), including fact verification and gap-driven content generation.
  • Work across the full AI surface of the product-agent design, orchestration, data storage, and tool integrations- taking features from concept to production in a fast-moving startup environment.
  • Evaluate agent performance for production readiness based on accuracy, latency, and token usage metrics.
  • Implemented a quick LLM call to ensure generated prompts met standards before being shown to users.

Results

  • Achieved a margin of reduction in hallucinations using an LLM-as-a-judge architecture for prompt validation.
PythonAgentic AIMulti-Agent OrchestrationWeb CrawlingLLM OrchestrationClaudeGPTGeminiTool UseDatabasesG-EvalLLM-as-a-judge

AI Engineer

InoTexelLahore, PKwork2026 – Present

Summary

IQPrompt - Prompt Optimization Platform for Developers Python • FastAPI • REST • RAG • OpenAI / Anthropic • Multi-tenant Backend SalesCoach - Real-Time Agentic Meeting Assistant Node.js • GraphQL • Neo4j • NAMS • LLM Orchestration

What I did

  • Architected IQPrompt, an API-first, multi-tenant platform that turns raw developer intent into personalized, retrieval-grounded prompts before any model call-defining the technical solution end to end, not just implementing a spec.
  • Designed the optimization pipeline as Intent Classification → RAG / knowledge retrieval → context injection → LLM → final prompt, a pattern directly reusable for agentic and RAG-based client solutions.
  • Defined multi-tenant architectural boundaries so customer knowledge stays isolated while a single scalable pipeline serves concurrent API consumers in production-a design decision validated through benchmark analysis for a technically skeptical stakeholder.
  • Led a four-stage G-Eval benchmark across 60 prompt cases and six domains to empirically justify architectural trade-offs, identifying metric artifacts and prioritizing follow-up validation experiments-demonstrating rigor in defending technical decisions to critical audiences.
  • Architected SalesCoach, an agentic, real-time meeting assistant designed so coaching suggestions are grounded in account history rather than only the live transcript.
  • Owned the end-to-end meeting-intelligence architecture: live transcript feeds a knowledge graph of clients, meetings, and organizations; NAMS retrieves historical client memory before the LLM generates a suggestion-an orchestration pattern applicable across agentic client solutions.
  • Exposed that memory through a shared GraphQL contract so the live assistant and backend systems stay architecturally aligned during live sessions.

Results

  • Result: real-time, agentic recommendations grounded in prior client interactions rather than a single conversation window.
PythonFastAPIRESTRAGOpenAIAnthropicMulti-tenant BackendNode.jsGraphQLNeo4jNAMSLLM OrchestrationG-EvalModel EvaluationBenchmarking

Associate ML Engineer

Taar ConsultingLahore, PKworkJul 2025 – Nov 2025

Summary

Key Responsibilities: • Delivered enterprise-scale data engineering over 97M+ transactions-production ETL, schema discipline, and failure-tolerant batch jobs for reporting that leadership could trust. • Designed automated ingestion and aggregation architecture into warehouse tables, replacing manual pipeline ops with reliable analytics infrastructure. • Hardened data reliability for production reporting so downstream analytics and ML consumers ran on consistent, large-scale transactional data.

What I did

  • Built production ETL and automated ingestion/aggregation pipelines for enterprise-scale data infrastructure
  • Focused on schema consistency, data reliability, and failure-tolerant processing to support analytics and ML workloads
  • Transformed large-scale raw transactional data into reliable datasets for downstream systems

Results

  • Worked with 97M+ transactional records
ETLData EngineeringAnalytics InfrastructureTraditional ML workflowsData Infrastructure

Skills

Technical

AI Agents
LLM Orchestration
Multi-Agent Systems
Agentic AI
Context Engineering
GraphQLGraphQL
RAG
Enterprise/Multi-Tenant Architecture
Neo4j
OpenAI
Neo4j Agent Memory Service (NAMS)
Knowledge Graphs
Anthropic
Prompt Engineering
PostgreSQLPostgreSQL
MCP
Intent Classification
Web Crawling & Data Extraction
Entity Relationships
Graph Traversal
Node.jsNode.js
FastAPIFastAPI
REST
GitGit
Next.jsNext.js
GitHubGitHub
OAuth
DockerDocker
AI Governance
VercelVercel
ReactReact
Auth0
TailwindTailwind
Railway
Remotion
Lemon Squeezy
Model Evaluation
Traditional ML workflows
G-Eval
LLM-as-a-judge

Languages

Python
TypeScript
SQL

Projects

Miless - AI-Native Relationship Intelligence

Architect / Developerpersonal
Architected Miless as an AI-native relationship intelligence platform for sales and interview workflows-meeting memory and client history need to be available live, not buried in CRM notes. Modeled clients, meetings, and organizations as a graph so the system retrieves context along real relationships instead of isolated rows. Justified graph-native architecture over traditional CRM tables, which flatten who-met-whom and prior discussion history, enabling the model to reason over organizational memory before it answers. Delivered GraphQL APIs for context retrieval in live and post-meeting flows; owned architecture, memory model, AI pipeline, and deployment from beta to production. Impact: recommendations and meeting intelligence reflect relationship history, not a single disconnected record.
TypeScriptNext.jsNode.jsGraphQLNeo4jLLM APIsDockerVercel

Grphly - Shared AI Memory Platform

Solution Architect / Lead Developerpersonal
Took Grphly from concept to production after identifying a recurring architectural gap: AI context scattered across Cursor, Claude, and ChatGPT, with no shared long-term memory between tools. Designed a collaborative, agent-agnostic memory platform-so teams stop re-explaining project context every time they switch tools. Selected MCP as the agent interface standard: a remote MCP server keeps memory access stable while Cursor, Claude, and custom agents remain interchangeable clients-an architectural choice built for extensibility across evolving AI ecosystems. Chose a graph store (with NAMS for session persistence) so entities and relationships stay traversable over time; PostgreSQL holds transactional metadata so concurrent writes stay consistent. Used GraphQL as the shared app/agent contract for memory operations, and OAuth/Auth0 to isolate namespaces per user for secure multi-tenant collaboration. Owned solution architecture, auth, AI pipelines, and cloud deployment end to end. Impact: one shared memory layer replaces repeated context switching across AI tools in day-to-day workflows.
TypeScriptNext.jsNode.jsGraphQLNeo4jNAMSPostgreSQLMCPOAuthAuth0DockerVercel

Education

BS Computer Science

FAST NUCESComputer ScienceAug 2021 – Jun 2025
Hire DaemReach out about a role, a contract or a conversation.For recruiters

Daem's twin is AI, it can make mistakes.

More people on Kanvis

Browse more peopleCreate your own Kanvis