AI/ML Team Lead & Head of AI focused on production LLM/RAG systems, multimodal pipelines, AI agents, KYC automation, and ML/data infrastructure. Also working on RL and harness engineering. I work from research and zero-to-one prototypes to production services, evaluation loops, cost/latency optimization, and team delivery.
What I work on
- Production AI / KYC
- Data infra / analytics
- Agents / RAG / LLM systems
- RL, Harness Engineering
- CV / multimodal / research
- Leadership / team delivery
| FOSS Project | Description | |
|---|---|---|
| 🧠 | ner | My production-grade llm-based simple plug-n-play ner implementation I use in a lot of my projects |
| 📖 | miRAGe | Repo with tons of experiments all around RAG and how different setups change the outcomes/benchmark results |
| 🧩 | agent-injector | MCP server that lets Claude delegate tasks to alternative LLM providers while keeping full agentic toolset and infrastructure of Claude Code |
| 💻 | rl-kyc-task-env | RL Environment with continuous scoring for one concrete kyc pipeline module subtask |
| 🦀 | openclaw-long-tasks | Long multi-step agent-crash resistant task scheduler/orchestrator for OpenClaw |
| 🐝 | agent-swarm | A CLI tool that allows you to process tasks by swarms of llms instead of only one of them (API can be exposed as a server for further integrations) |
| 🔎 | mcp-probe | MCP server automated testing tool |
| 🔍 | perplexity-mcp | MCP server bringing Perplexity's web search and deep research into any AI workflow (Cursor, Claude, Codex, OpenClaw etc.) |
| ♜ | age_of_prikol | Age of War inspired game developed in Zig and Raylib to get a feeling for them (also I love Age of War a lot) |
| 🖼️ | jeeja-image-labeller | A simple local image labeller for ML datasets built in C#, supports YOLO export format |
