Jiale M.

Key Skills

Software

No software listed

Top Subject Matter

Cloud resource scheduling

agentic LLM systems

AI/ML experimentation

Top Data Types

Text

Audio

Document

Top Task Types

Question Answering

Text Generation

RLHF

Prompt + Response Writing (SFT)

Freelancer Overview

LLM Evaluator / AI Trainer – Multi-Agent Cloud Resource Scheduler Project. Brings 1+ years of professional experience across complex professional workflows, research, and quality-focused execution. Core strengths include Internal and Proprietary Tooling. Education includes Master of Science, The University of Edinburgh (2025) and Bachelor of Science, Xi’an Jiaotong-Liverpool University (2025). AI-training focus includes data types such as Text and labeling workflows including Evaluation and Rating.

Labeling Experience

LLM Evaluator / AI Trainer – Multi-Agent Cloud Resource Scheduler Project

Text

Led the evaluation and structured validation of LLM agentic outputs for planning and scheduling tasks in a multi-agent cloud resource management project. Focused on schema-constrained tool-call outputs, parse-success monitoring, hallucination detection, and deterministic fallback implementations. Evaluated model reasoning by transforming server state vectors into natural-language prompts and measuring compliance with hard constraints. • Designed and implemented JSON schema-based tool-call evaluations for LLM agents. • Benchmarked LLM outputs against heuristic and rule-based baselines. • Created self-distilled SFT data from decision traces for fine-tuning. • Built reproducible test pipelines to ensure labeling and evaluation quality.

2026 - 2026

Agent Evaluation Specialist / Lead Engineer – AIroute Project

Text

Directed agentic task, planning, and output schema evaluation in a local trip planning LLM agent project. Implemented structured intent schema checks, schema-constrained response validation, and route repair/revalidation cycles based on LLM-generated plans. Led the review, grading, and auditable scoring of LLM plans and replanning behaviors under user constraints. • Developed 10-dimensional scoring breakdowns for agent recommendations. • Validated compliance of plans to user-specified constraints and preferences. • Recorded and explained revisions and trade-offs in model behavior. • Supported QA through robust test pipelines and feedback loops.

2025 - 2026

LLM Evaluation Contributor – Intelligent Cloud Management System Research

Text

Developed prompt pipelines and simulated decision breakdowns for LLM-based agent evaluation in intelligent cloud management research. Compared LLM-driven decisions to traditional rule-based and heuristic scheduling outcomes, providing detailed feedback and constraint validation on model selections. Focused on evaluating reliability under uncertainty, schema compliance, and structured natural-language prompts for model improvement. • Integrated LangChain/Ollama and NetLogo for labeling and simulation workflows. • Validated structured output reliability under resource constraint scenarios. • Designed and tracked prompt-driven evaluation data for agent actions. • Provided feedback on model error cases, output validation, and fallback logic.

2024 - 2025

AI Evaluation Contributor – LSTM-DQN Robotic Fish Navigation

Text

Participated in iterative evaluation and robustness testing of decision-making models for LSTM-DQN robotic navigation in simulated environments. Designed and executed data-driven cross-validation and ablation studies, performing review of reward-based model adjustments and navigation outcome analysis. Interpreted agent behaviors and provided feedback on simulation settings and model outputs to refine both the training and evaluation process. • Carried out sequential decision evaluation using OpenAI Gym and ML frameworks. • Systematically compared and reported model robustness under varying assumptions. • Collaborated with technical domain experts to improve simulation and labeling criteria. • Supported research publication with thorough experimental documentation.

2023 - 2024

Education

Work History