Available for AI Engineering Roles

Engineering Autonomous AI Agents & Scalable Systems.

Hi, I'm Kirollos Henry. I specialize in multi-agent orchestration with LangGraph & LangChain, human-in-the-loop agent workflows, Text-to-SQL pipelines, and advanced Active RAG systems.

Featured AI Implementations

06 Total Builds
LangGraph Multi-Agent Research

Research-AI

LangGraph Human-in-the-Loop Multi-Agent

Autonomous deep-research assistant built as a LangGraph state machine. A clarifier agent pauses the graph to ask the user follow-up questions when a query is vague, a planner splits it into sub-questions, parallel researcher subgraphs search the web with Tavily, and a writer agent compiles a source-cited markdown report.

  • HITL clarification via interrupt() & checkpointed resume
  • Parallel researcher fan-out with the Send API
  • Tool-calling subgraphs with capped web-search loops
  • Pydantic structured output & live streamed progress
LangGraphGoogle GeminiTavilyStreamlitDocker
Repository & Docs
Text-to-SQL Orchestration

AskDB-AI

Production-grade multi-agent orchestrator for natural language database interaction. Features a decoupled pipeline for intent disambiguation, dynamic schema pruning, and self-correcting SQL generation loops, ensuring high precision and context efficiency.

  • Multi-agent orchestration pipeline
  • Dynamic schema pruning & context optimization
  • Autonomous self-correction & retry loops
Google GeminiText-to-SQLFunction CallingMulti-Agent
Repository & Docs
LangGraph Multi-Agent Workflow · v2.0

OfficeAssistant-AI

LangGraph LangChain Human-in-the-Loop Multi-Agent

Email orchestration assistant rebuilt on LangChain + LangGraph. A planner agent breaks each request into dependent sub-tasks and fans them out with Send to specialised inbox, explain and compose sub-agents, running independent tasks in parallel. Every outgoing email pauses the graph with interrupt() for human approval before it is sent.

  • HITL approval gate: approve, edit, request changes or cancel
  • Planner → dependency-aware parallel sub-agents via Send
  • Email explainer: intent, deadlines, action items & risk flags
  • Registry-driven sub-agents & tools with checkpointed threads
LangGraphLangChainGoogle GeminiPydanticGmail IMAP/SMTPStreamlit
Repository & Docs
Active Retrieval RAG

CodeProbe-AI

Repository comprehension engine using active vector retrieval and semantic embeddings. Automatically indexes project hierarchies, executes function calls to inspect syntax and files, and enables grounded natural language code Q&A.

  • Dynamic multi-vector code retrieval
  • Multi-file context expansion & AST inspection
  • Repository-grounded Q&A architecture
Vector DBText EmbeddingsActive RAGCodebase Parsing
Repository & Docs

DocuSense-AI Demo

Video demo slot ready
Document Intelligence

DocuSense-AI

Specialized analytical pipeline for complex PDF documentation. Uses active retrieval, semantic chunking, and tool-use function calls to deliver citation-accurate answers across voluminous multi-page reports.

  • Dynamic chunking & document parsing
  • Vector similarity semantic search
  • Citation-grounded question answering
RAG PipelineEmbeddingsPDF AnalysisPython
Repository & Docs

AI-Route-Planner Demo

Video demo slot ready
Full-Stack System

AI-Route-Planner

Autonomous travel orchestration platform backed by live web discovery agents. Uncovers unlisted regional transit options and designs comprehensive multi-modal itineraries via an asynchronous backend.

  • Live web grounding for public transit
  • High-concurrency FastAPI architecture
  • PostgreSQL relational persistence layer
FastAPIReactPostgreSQLWeb Search Agent
Repository & Docs