Suparna Chakraborty

Key Skills

Software

Lionbridge

Top Subject Matter

LLM Fine-tuning & Evaluation (South Asian Languages)

AI Safety & RLHF (Indic/Regional Languages)

Legal Services & Contract Review

Top Data Types

Text

Document

Top Task Types

Fine Tuning

RLHF

Freelancer Overview

Human-in-the-Loop LLM Evaluator & Quality Lead. Brings 10+ years of professional experience across legal operations, contract review, compliance, and structured analysis. Core strengths include Lionbridge. Education includes Postgraduate Diploma, University of Michigan (2024) and Postgraduate Diploma, Balridge University (2011). AI-training focus includes data types such as Text and labeling workflows including Evaluation and Rating.

Labeling Experience

Data annotation and labeling

TextRLHF

2 Lacs of words

2 Lacs of words

2009 - 2025

LLM Fine-Tuning and Evaluation Lead (Google Research & DeepMind)

LionbridgeTextFine Tuning

Developed and implemented a systematic human-in-the-loop (HITL) evaluation framework for assessing generative model outputs in South Asian languages. Led a team to evaluate factuality and semantic alignment, resulting in measurable reductions in model hallucination rates prior to deployment. Architected and managed the linguistic quality benchmarking process, including building a quality scorecard for annotation consistency. • Evaluated responses across 10+ South Asian dialects against English source prompts. • Reduced factual inaccuracies and improved model output quality. • Led multi-phase manual rating of LLM-generated text. • Established benchmarks for multi-lingual model performance.

Developed and implemented a systematic human-in-the-loop (HITL) evaluation framework for assessing generative model outputs in South Asian languages. Led a team to evaluate factuality and semantic alignment, resulting in measurable reductions in model hallucination rates prior to deployment. Architected and managed the linguistic quality benchmarking process, including building a quality scorecard for annotation consistency. • Evaluated responses across 10+ South Asian dialects against English source prompts. • Reduced factual inaccuracies and improved model output quality. • Led multi-phase manual rating of LLM-generated text. • Established benchmarks for multi-lingual model performance.

2021 - 2024

RLHF Data Labeling & AI Safety Strategist (Meta Labs)

TextRLHF

Managed high-precision labeling and curation of training data for Reinforcement Learning from Human Feedback (RLHF) projects with Meta Labs. Focused on identifying and mitigating subtle socio-cultural and gender biases within regional language datasets. Authored safety-alignment guidelines which were adopted into standard pipelines. • Oversaw annotation of over 1M tokens for RLHF training. • Designed workflows for capturing the nuances of polite/honorific forms in Indic languages. • Ensured all data labeling met stringent ethical and accuracy standards. • Facilitated integration of labeled datasets for regional language deployment.

Managed high-precision labeling and curation of training data for Reinforcement Learning from Human Feedback (RLHF) projects with Meta Labs. Focused on identifying and mitigating subtle socio-cultural and gender biases within regional language datasets. Authored safety-alignment guidelines which were adopted into standard pipelines. • Oversaw annotation of over 1M tokens for RLHF training. • Designed workflows for capturing the nuances of polite/honorific forms in Indic languages. • Ensured all data labeling met stringent ethical and accuracy standards. • Facilitated integration of labeled datasets for regional language deployment.

2020 - 2023

Education

Work History