Focused on making AI reliable, grounded, and deployable in the real world.

"A model is only useful when it's reliable, grounded, and deployable."

• Building RAG systems with an emphasis on grounding, retrieval quality, and reducing hallucinations
• Developing LLM evaluation workflows to measure factuality, relevance, and response quality
• Engineering end-to-end ML systems across data processing, model inference, APIs, and deployment
• Building MCP servers and AI tooling that allow LLMs to interact with real-world tools and systems
• Working with LangGraph and agentic workflows for multi-step AI applications
• Deploying AI applications using Docker, cloud infrastructure, and production-oriented backend architectures

I founded and maintain AynOps — an open-source MCP server for reconnaissance and security analysis. As a maintainer, I:

As a maintainer, I:

• Design and open bug reports and enhancement issues
• Investigate bugs, identify edge cases, and improve issue reports
• Review and test community pull requests
• Maintain and expand the project's test suite
• Improve documentation and contributor experience
• Help contributors understand issues and make focused contributions through discussions
• Maintain the project's architecture and MCP tools
• Maintain CI workflows and automated checks
• Track project development through releases and changes

• LogLense — Contributor. Added 8+ new CLI subcommands across repository scanning, log parsing, filtering, file statistics, language detection, Git analysis, and monitoring workflows. Improved command-line usability with new options, output handling, and UI/UX enhancements, while adding and updating tests to ensure the new functionality remains reliable.
• py-simple-wrap — Collaborator. Contributed utility functions and test coverage to improve the project's functionality and reliability. Worked on edge cases and validation through automated tests, helping strengthen the library's behavior and maintainability.
• anthropics/modelcontextprotocol/inspector — Investigated and reported a JSON editor serialization bug affecting nullable parameters, where null values could be transformed into incorrectly escaped JSON through repeated editing. Documented the reproduction steps, expected vs. actual behavior, and impact to help maintainers reproduce and resolve the issue.
• Other open-source contributions — 50+ PRs across 20+ repositories, spanning Docker, CI/CD, testing, backend fixes, and developer tooling. Explore my pull requests →

• RizqAI — Multi-agent financial analysis platform using LangGraph, LangChain, Langsmith and Gemini, with specialized agents for planning, research, risk assessment, adversarial debate, and investment thesis generation, backed by stateful workflows and SQLite checkpointing.
• RoasterBro - AI-powered CLI for repository intelligence and developer interrogation, combining multi-signal codebase analysis, Git insights, dependency analysis, and LLM-powered 3-round ragebait reasoning with support for local Ollama and cloud models.
• ClipForge — AI short-clip generator using LangGraph, FFmpeg, Whisper, and Groq, with asynchronous video processing and real-time rendering progress.
• Grounded LLM QA — RAG-based QA system using LLaMA-3.1, LangChain, and ChromaDB/FAISS, with LLM-as-a-judge evaluation; reduced hallucination rates from 60% to 15% through retrieval and context optimization.
• Fashion Recommender System — Visual recommendation system using a fine-tuned ResNet50, deployed as a Dockerized inference service on AWS EC2 with Nginx and Gunicorn, achieving sub-second inference latency.

AI & ML

Data & NLP

Deployment & Infrastructure

⭐ If something in my repositories helped you, consider giving it a star — it means a lot!