Ulas Ahmet Kirac — Data Science & AI Student
I build LLM-powered agents and natural-language interfaces — turning real-world data into things people actually use.
I'm a BSc Data Science & Artificial Intelligence student at Leiden University (expected 2027), focused on building practical AI systems — from LLM-powered agents to natural-language interfaces over real company data.
I've worked across the modern AI stack: shipping an Agent-to-Agent integration for Unilever's FoodsGPT, leading a team that built an embodied virtual assistant in Unity, and building a pricing tool that cut ~85% of manual work at Borunet. I like ambiguous problems, agentic systems, and getting research-grade ideas into production.
Skills
- Core Languages & Tools: Python, Java, JavaScript, SQL, Git, Docker, Unity, SQLite
- Machine Learning & Data: PyTorch, Scikit-learn, Pandas, NumPy, OpenCV, MediaPipe, Plotly, Dash, Jupyter
- LLMs & Agentic Systems: LangGraph, LangChain, Agent-to-Agent (A2A), RAG, MCP, Anthropic & OpenAI APIs, Streamlit, Flask
- Spoken Languages: English — C2, Turkish — C2, Dutch — A2
Projects
- “Confetti” — Virtual Birthday Assistant — Team lead for a 4-person project building an embodied virtual birthday assistant focused on emotional and social interaction. Built in Unity with the ConvAI plugin, integrating the ConvAI API and ReadyPlayerMe for real-time 3D-avatar conversation. Validated with 10+ users, reaching 90% positive feedback on interaction quality and conversational flow.
- FoodsGPT — Agent-to-Agent Integration — Designed and implemented an Agent-to-Agent (A2A) integration with LangGraph at Unilever, connecting the FoodsGPT natural-language interface to the Chef Buddy recipe-intelligence agent — expanding the tools and data sources available to Foods R&D product developers. Evaluated retrieval quality through user interviews, scenario testing, and relevance analysis.
- Landed-Cost & Margin Calculator — Built a landed-cost & margin calculator (FOB/CIF, FX, freight, customs) that produced quote-ready prices in seconds, eliminating ~85% of previously manual calculation work. Applied SQL and pandas against the company's internal database and processed 10,000+ transaction records for quarterly margin analysis.
Experience
- Teaching Assistant at Leiden University (Sep 2026 — Present) — Mentor student teams through their end-of-semester Human-Agent Interaction research projects (Prof. Joost Broekens), advising on agent design, multimodal interaction, and experimental evaluation with real users. Support prototype development across virtual agents (Unity, ConvAI) and physical robots (NAO, AlphaMini), and grade deliverables with structured feedback on implementation, methodology, and academic writing.
- Data Science, R&D Intern at Unilever (Mar 2026 — Sep 2026) — Contributed to FoodsGPT, a natural-language AI interface connecting Foods R&D product developers to internal systems used across Unilever's regional offices. Built an Agent-to-Agent integration with LangGraph and evaluated retrieval quality and end-user value through user interviews, scenario testing, and relevance analysis.
- Data Scientist, Part-Time at Borunet (Aug 2025 — Jan 2026) — Built a landed-cost & margin calculator that produced quote-ready prices in seconds, eliminating ~85% of previously manual work. Applied SQL and pandas against the company's internal database and processed 10,000+ transaction records to support quarterly margin analysis.
Contact
Email: okululas@gmail.com