Ефим Х.

Key Skills

Software

Appen

Other

Internal/Proprietary Tooling

Don't disclose

Top Subject Matter

Artificial Intelligence & Technology: Proficient in Python programming

Large Language Model (LLM) workflows

Retrieval-Augmented Generation (RAG) system testing.

Top Data Types

Image

Text

Computer Code Programming

Document

Top Task Types

Evaluation/Rating

Computer Programming/Coding

Text Generation

Text Summarization

Entity (NER) Classification

Question Answering

RLHF

Red Teaming

Freelancer Overview

A highly analytical AI Trainer and and Data Specialist with over 3 years of handson experience in fine-tuning, aligning, and evaluating Large Language Models at scale within a major AI ecosystem (Yandex). Combining a solid software engineering foundation (Python, SQL, Flutter) with deep expertise in RLHF, SFT, and AI Safety. Specialized in static code verification, search relevance engineering (SERP), and establishing ethical guardrails for global AI models.

Labeling Experience

Generation, Prompt Engineering, Yandex (

TextRLHFEntity (NER) Classification

No description provided.

2023 - Present

Model evaluation for SERP/search relevance — YandexGPT/AI training

Don't discloseText

Conducted technical AI evaluations focused on search relevance and SERP quality. Spearheaded side-by-side benchmarking (Yandex vs. Google) to compare performance for search-integrated AI models. Performed relevance verification through systematic evaluation of outputs and ranking behavior. • Compared SERP output quality across engines • Performed relevance verification and rating • Supported benchmarking for search-integrated AI • Used inspection and validation workflows to assess model output relevance

2023 - Present

AI Safety & QA within Lead AI-Trainer role — YandexGPT

Don't discloseTextRed Teaming

Audited LLM responses for ethical compliance and safety guideline adherence as part of AI safety and QA practices. Implemented defensive evaluation strategies to mitigate adversarial prompts, jailbreak attempts, and prohibited content. Performed factual and hallucination checks to improve reliability and trustworthiness of outputs. • Reviewed responses for ethical and safety compliance • Mitigated adversarial prompting and jailbreak attempts • Performed hallucination and fact-checking validation • Reduced unsafe/prohibited content in model outputs

2023 - Present

Lead AI-Trainer / Technical Content Specialist — Yandex (Model Alignment / AI training)

Don't discloseTextRLHF

Served as Lead AI-Trainer / Technical Content Specialist for YandexGPT, performing RLHF and SFT-related training workflows. Led evaluation, annotation, and alignment processes to improve model behavior using high-quality prompt-response gold standard datasets. Conducted defensive quality work to strengthen human-preference ranking and overall model accuracy. • Evaluated and annotated training examples for alignment • Built/curated prompt-response gold standard datasets • Supported model training via RLHF/SFT workflows • Improved accuracy through preference-aware evaluation

2023 - Present

Lead AI Trainer, YandexGPT — RLHF, evaluation, annotation, and LLM alignment

TextRLHF

Provided reinforcement learning from human feedback workflows for YandexGPT to improve response accuracy and safety. Created high-quality prompt-response datasets and established human-preference ranking criteria for model alignment and evaluation. Performed evaluation and annotation activities as part of the alignment pipeline to support LLM training iterations. • Curated prompt-response training data for preference modeling • Designed human preference ranking/criteria used in RLHF • Conducted evaluation to measure accuracy and safety benchmarks • Implemented hallucination detection and fact-checking validation frameworks

2023 - Present

Education

Work History