Ivan S.

Key Skills

Software

No software listed

Top Subject Matter

Physics (mechanics, electromagnetism, quantum, general physics)

Mathematics (calculus, linear algebra, mathematical reasoning)

Python programming and scientific computing

Top Data Types

Text

Computer Code Programming

Document

Top Task Types

Prompt + Response Writing (SFT)

Fine-tuning

RLHF

Computer Programming/Coding

Question Answering

Evaluation/Rating

Text Generation

Function Calling

Text Summarization

Red Teaming

Freelancer Overview

Physics student at Lomonosov Moscow State University (MSU), Faculty of Physics. Solid foundation in mathematics, theoretical and experimental physics, and Python programming for scientific computing. Founder of an early-stage AI project: an LLM-based agent for STEM researchers and PhD students that performs deep literature review across arXiv, eLibrary, Math-Net.ru, and CyberLeninka, builds citation graphs, and produces IMRaD/LaTeX-formatted output. This gives me hands-on experience with LLM prompting, RAG pipelines, evaluating model responses on scientific content, and working with academic documents in Russian and English. Strengths for AI training tasks: — Evaluating and ranking LLM responses on physics, math, and Python coding problems (RLHF, preference comparison) — Writing high-quality prompt + response pairs (SFT) for STEM reasoning — Reviewing Python code for correctness and edge cases — Labeling scientific documents and LaTeX content — Function calling and agent-style task evaluation Languages: Native Russian, conversational English for technical content. Comfortable with both English- and Russian-language LLM data tasks.

Labeling Experience

Ongoing AI/NLP relevant work — fine-tuning and RAG pipeline development

Fine-tuning

Working on training and optimization of NLP models, including BERT fine-tuning for Russian language tasks. Also developing classical NLP baselines and building retrieval-augmented generation (RAG) pipelines for applied use cases. • Fine-tuning ruBERT (rubert-tiny2) for Russian NLP. • Implementing TF-IDF and embedding-based classical NLP components. • Applying quantization and ONNX export for inference optimization. • Developing RAG pipelines and related retrieval components.

2025 - Present

Python Developer (R&D) — LLM agent engineering and model analysis (photonic computing)

Prompt + Response Writing (SFT)

Developed and shipped an LLM agent that automates photonic circuit design, requiring prompt/context engineering and multi-step reasoning chains. Performed model graph decomposition and inference profiling to optimize agent and model behavior for specialized hardware. • Built a production LangGraph agent for Lumerical CAD integration. • Engineered prompts and context for reliable tool-using behavior. • Conducted PyTorch FX graph decomposition and performance analysis. • Worked on ongoing NLP/LLM training and optimization tasks including BERT fine-tuning and RAG pipelines.

2024 - 2026

Education

Work History