Bhaskar Joshi

Key Skills

Software

Other

Mindrift

Appen

Labelbox

Label Studio

Micro1

Toloka

Telus

Top Subject Matter

AI language models

Rlhf Domain Expertise

Multimodal AI

Top Data Types

Text

Document

Video

Top Task Types

Question Answering

Prompt + Response Writing (SFT)

Transcription

Fine-tuning

RLHF

Evaluation/Rating

Freelancer Overview

I am a civil engineer and structural engineering postgraduate with strong expertise in mathematics, analytical reasoning, and research. My academic foundation is complemented by hands-on experience in teaching, engineering supervision, and content development for licensing examinations. I approach problems methodically, emphasizing precision, clarity, and practical applicability across engineering and quantitative domains. I have worked extensively in AI training and evaluation, contributing to large language model development through prompt design, quality analysis, and reinforcement learning workflows. In roles involving leadership and quality assurance, I have guided teams, refined model reasoning, and ensured high standards of accuracy and consistency. My work reflects a focus on structured thinking, effective communication, and aligning technical systems with reliable, human-centered outcomes.

Labeling Experience

AI Trainer (Pareto.ai — Howlite/Sapphire Projects)

OtherText

As an AI Trainer at Pareto.ai (Howlite and Sapphire projects), I reviewed and validated labeled solution data for mathematics and engineering problems at university and competition levels. I provided step-by-step accuracy and quality checks, utilizing rubrics and QA protocols for error detection. I further contributed by designing prompts for agentic AI, guiding it in workflow automation and performance improvement. • Validated labeled academic problem solutions in algebra, calculus, combinatorics, and more. • Applied quality assurance methods to detect logical flaws and inconsistencies. • Designed and graded rubrics for complex mathematical and engineering tasks. • Created prompts and real-time feedback mechanisms for AI agent guidance.

As an AI Trainer at Pareto.ai (Howlite and Sapphire projects), I reviewed and validated labeled solution data for mathematics and engineering problems at university and competition levels. I provided step-by-step accuracy and quality checks, utilizing rubrics and QA protocols for error detection. I further contributed by designing prompts for agentic AI, guiding it in workflow automation and performance improvement. • Validated labeled academic problem solutions in algebra, calculus, combinatorics, and more. • Applied quality assurance methods to detect logical flaws and inconsistencies. • Designed and graded rubrics for complex mathematical and engineering tasks. • Created prompts and real-time feedback mechanisms for AI agent guidance.

2025 - Present

AI Domain Expert (Freelance)

MindriftTextQuestion Answering

As an AI Domain Expert (Freelance) at Mindrift, I developed complex question-answer pairs and constructed scoring rubrics for evaluating AI reasoning. I identified and manually corrected AI mistakes to improve answer quality and reliability. My efforts supported reinforcement learning and prompt optimization for further model training refinement. • Authored high-level academic and reasoning challenge questions for AI evaluation. • Built detailed model evaluation rubrics for consistent scoring. • Conducted error analysis and provided corrective feedback loops. • Enhanced model performance through targeted prompt design and data contributions.

As an AI Domain Expert (Freelance) at Mindrift, I developed complex question-answer pairs and constructed scoring rubrics for evaluating AI reasoning. I identified and manually corrected AI mistakes to improve answer quality and reliability. My efforts supported reinforcement learning and prompt optimization for further model training refinement. • Authored high-level academic and reasoning challenge questions for AI evaluation. • Built detailed model evaluation rubrics for consistent scoring. • Conducted error analysis and provided corrective feedback loops. • Enhanced model performance through targeted prompt design and data contributions.

2025 - 2025

Quality Analyst/Pod Lead (Meta Multimodal Projects)

OtherVideoRLHF

As a Quality Analyst/Pod Lead in the Meta Multimodal SFT and DPO projects, I was responsible for quality assurance of labeled AI data and leadership in evaluation workflows. I ensured team productivity, adherence to guidelines, and improvement of project standards through reviews and feedback. I contributed to several related projects, focusing on mathematical and logical reasoning feedback for AI outputs. • Managed QA processes and mentored labeling team members. • Oversaw annotation and evaluation quality for text and math prompt responses. • Supported improvement on Math Rewrites, Graph Rewrites, Meta Reasoning, and Amazon Reasoning Projects. • Participated actively in SFT and DPO annotation phases, providing evaluative feedback.

As a Quality Analyst/Pod Lead in the Meta Multimodal SFT and DPO projects, I was responsible for quality assurance of labeled AI data and leadership in evaluation workflows. I ensured team productivity, adherence to guidelines, and improvement of project standards through reviews and feedback. I contributed to several related projects, focusing on mathematical and logical reasoning feedback for AI outputs. • Managed QA processes and mentored labeling team members. • Oversaw annotation and evaluation quality for text and math prompt responses. • Supported improvement on Math Rewrites, Graph Rewrites, Meta Reasoning, and Amazon Reasoning Projects. • Participated actively in SFT and DPO annotation phases, providing evaluative feedback.

2024 - 2025

LLM Trainer—Research Analyst (Apple RLHF Project)

OtherText

As an LLM Trainer—Research Analyst on the Apple RLHF project, I evaluated and provided detailed feedback on model-generated responses. My work guided the model training process by shaping AI learning through targeted critique and reinforcement. I identified weaknesses and inconsistencies in the model’s logic, directly contributing to its improvement. • Conducted comprehensive evaluations of AI responses to user prompts. • Provided specific actionable feedback to enhance model accuracy and safety. • Collaborated with team members to spot areas for reasoning improvement. • Played an active role in the reinforcement learning loop by rating and annotating model outputs.

As an LLM Trainer—Research Analyst on the Apple RLHF project, I evaluated and provided detailed feedback on model-generated responses. My work guided the model training process by shaping AI learning through targeted critique and reinforcement. I identified weaknesses and inconsistencies in the model’s logic, directly contributing to its improvement. • Conducted comprehensive evaluations of AI responses to user prompts. • Provided specific actionable feedback to enhance model accuracy and safety. • Collaborated with team members to spot areas for reasoning improvement. • Played an active role in the reinforcement learning loop by rating and annotating model outputs.

2024 - 2024

Education

Work History