Hardik K.
Team Lead - AI Training
Experience
Jun 2025 - Present
6 monthsUnited States
AI Developer - LLM Trainer (Amazon)
Turing
- Trained Amazon’s proprietary AI models to interpret and reason over graphs, scientific visuals, and analytical plots sourced from research papers and technical documents.
- Created complex problem sets requiring visual approximation, multi-step analysis, trend detection, and value inference—mirroring real-world analytical tasks and exam-style reasoning.
- Benchmarked Amazon’s models against industry leaders (GPT-4, Claude, Gemini), fine-tuned prompts, and improved performance through error analysis and custom visual reasoning workflows.
- Collaborated with Amazon’s research team to integrate multimodal training feedback and improve model grounding across academic and real-world visual datasets.
- Led and reviewed work from junior AI trainers and new joiners—refining data annotation, visual comprehension questions, and maintaining quality across multimodal datasets.
Jan 2025 - Jun 2025
6 monthsUnited States
AI Developer - LLM Trainer (Google)
Turing
- Fine-tuned LLMs using Supervised Fine-Tuning (SFT) on advanced mathematics problems from JEE and Olympiad-level datasets, optimizing both structured reasoning and final-answer accuracy.
- Designed clear, human-readable solutions by combining advanced prompt engineering with deep mathematical research, leveraging models such as GPT-4o and O-1 Preview to improve clarity, precision, and step-by-step logic.
- Led AI data training workflows—created, labelled, and audited high-quality datasets using CVAT and other annotation tools to enhance model comprehension of mathematical expressions and logic.
- Reviewed and refined outputs from junior engineers and new joiners, ensuring consistency, correctness, and alignment with pedagogical and project standards.
Feb 2024 - Jan 2025
1 yearUnited States
Prompt Engineer - AI Trainer
Outlier AI
- Designed, tested, and refined prompts for AI models across projects, including Kepler-RLHF, Psychic – Kadabra, Cypher-RLHF, and Bee-SFT-Multilingual, ensuring high accuracy in STEM, coding, general science, and knowledge-based problem-solving.
- Developed optimized prompt structures to enhance AI-generated responses, improving logical reasoning, clarity, and contextual relevance across diverse academic and practical domains.
- Evaluated AI models for Indian English (IN), refining linguistic precision and cultural context to support inclusive and accurate user interactions.
- Conducted systematic performance analyses using RLHF, SFT, reward modelling, and advanced prompting techniques to improve model consistency, response quality, and user alignment.
- Collaborated with cross-functional teams to scale and enhance prompt-based training workflows, contributing to robust advancements in AI-driven cognitive capabilities.
Jan 2024 - Jan 2025
1 year 1 monthChandigarh, India
Data Scientist
Medlr
- Developed the RapidFuzz technology algorithm using NLP, reducing medication errors by 30% and processing data from 697,620 medicines across top pharmacies for precision and efficiency.
- Automated web scraping for pharmaceutical data from platforms like Tata1mg, Pharmeasy, and Netmeds, integrating it into ElasticSearch to enhance data-driven insights.
- Designed a scheduling system for periodic data updates, expanding the dataset by 30% year-on-year while tracking price changes.
- Leveraged multithreading to accelerate data extraction and improve throughput, driving greater efficiency and productivity.
May 2023 - Dec 2023
8 monthsUnited States
Prompt Engineer - AI Tutor
Soul AI
- Developed and fine-tuned interactive prompts for large language models (LLMs), improving response accuracy and relevance by 25%, leading to enhanced AI-driven interactions.
- Collaborated with subject matter experts (SMEs) to integrate industry-specific terminology, boosting prompt effectiveness and user engagement by 15%.
- Conducted data annotation and labelling for text-based NLP tasks and image datasets, ensuring high-quality training data for AI models.
- Monitored prompt performance in real-world scenarios, analysed feedback, and iteratively optimised AI-generated responses, streamlining the data collection and decision-making process.
Summary
I am an AI Developer, AI Trainer/Tutor, and Data Scientist with hands-on expertise in Python, SQL, and advanced data visualization tools such as Power BI and Tableau. Currently, I serve as a Team Lead – AI Training at Turing (OpenAI projects), where I manage 25+ annotators and pod leads across multimodal, robotics, and reasoning-based model training initiatives, focusing on LLM fine-tuning, prompt engineering, and large-scale data delivery leadership. My prior work with Turing – Google and Turing – Amazon involved developing evaluation workflows, reasoning datasets, and fine-tuning models for analytical and domain-specific problem-solving. With cross-functional experience spanning machine learning, NLP, and data analysis, along with contributions to Medlr, Soul AI, and Outlier AI, I specialize in building automation pipelines, improving model reasoning through structured evaluation, and translating complex data into actionable, human-readable insights.
Languages
English
NativeHindi
NativeEducation
Thapar Institute Of Engineering And Technology
BE/BTech · Computer Science · Patiala, India
N.K. Bagrodia Public School
X · Delhi, India · 9 CGPA
N.K. Bagrodia Public School
XII · Delhi, India · 90%
Certifications & licenses
AI Engineer
Udemy
AWS Developer Associate Certification
Data Science and ML
Udemy
Google Data Analytics
Microsoft Excel: Excel from Beginner to Advanced
Microsoft Power BI for Business, Finance and Data Analysis
Need a freelancer? Find your match in seconds.
Try FRATCH GPT More actions
Similar Freelancers
Discover other experts with similar qualifications and experience
- English
- Deutsch
2025 © FRATCH.IO GmbH. All rights reserved.