AI engineer building deterministic agent systems, LLM infrastructure, and AI-powered applications.
My work focuses on making AI systems reliable, traceable, and reproducible. I design multi-agent architectures, deterministic execution engines, and real-world AI applications that convert unstructured input into structured, actionable outputs.
Deterministic multi-agent system that analyzes CI failures and identifies root causes using structured evidence and reproducible reasoning.
Key features
- Deterministic failure graph construction
- Ranked root-cause candidates with explicit scoring
- Evidence-based confidence scoring
- Structured outputs (
ci-rca.json,ci-rca.md) - Full traceability and replayable execution
Focus areas
- Multi-agent orchestration
- Deterministic reasoning pipelines
- CI observability and debugging automation
Execution engine and SDK for building reproducible LLM workflows with explicit state, structured traces, and deterministic guarantees.
Key features
- Explicit workflow state machines
- Structured execution traces
- Deterministic workflow execution
- Replayable runs and debugging support
- Infrastructure for reliable LLM systems
Focus areas
- LLM infrastructure
- Workflow orchestration
- Execution engines for AI systems
AI-powered iOS application that understands screen content and converts it into structured, actionable workflows using LLMs.
ScreenFlow enables users to take screenshots and instantly perform intelligent actions such as extracting structured data, summarizing content, generating replies, or triggering workflows.
Key features
- Screenshot-based AI understanding
- Structured extraction from visual input
- Context-aware action generation
- LLM-powered reasoning over screen content
- AI-driven workflow automation
Focus areas
- AI-powered mobile applications
- Human-AI interaction systems
- LLM-powered structured extraction
- Real-world AI usability
Execution engine for structured reasoning pipelines where every step is explicit, traceable, and replayable.
Key features
- Multi-step reasoning pipelines
- Explicit intermediate state representation
- Deterministic execution guarantees
- Trace generation and replay
- Structured reasoning outputs
Focus areas
- Agent reasoning systems
- Deterministic execution engines
- Traceable AI pipelines
Agent Systems
- Multi-agent architectures
- Deterministic reasoning systems
- Agent orchestration
- Execution engines
LLM Infrastructure
- Workflow orchestration
- Trace and observability tooling
- Structured execution pipelines
- Evaluation-aware system design
AI Applications
- LLM-powered applications
- Screenshot and visual understanding workflows
- Structured extraction from unstructured input
- AI-driven automation tools
Backend Systems
- Python backend development
- FastAPI services
- State machines and execution pipelines
- Structured logging and tracing
Languages
- Python
- Swift
- TypeScript
- Rust
AI / LLM
- OpenAI API
- LangChain
- LangGraph
- FAISS
Backend
- FastAPI
- REST APIs
- Execution pipelines
- Structured logging and tracing
Mobile
- iOS
- Swift
- AI-powered mobile workflows
Databases
- PostgreSQL
- SQLite
- SQL
Infrastructure
- GitHub Actions
- CI/CD systems
- Structured system design
- Deterministic execution over non-reproducible behavior
- Explicit state over implicit state
- Traceability and replayability
- Evaluation-driven system design
- Reliability as a core system property
Explore my repositories to see implementations of deterministic agent systems, execution engines, and AI-powered applications:
https://github.com/ibrahim1023