Saurabh M.-Doctoral Candidate • IIT Gandhinagar • AI Evaluator • Indic Specialist

I operate at the unique intersection of STEM precision and deep humanities scholarship:

• Doctoral Researcher at IIT Gandhinagar specializing in South Asian Archaeology, History, and Manuscriptology.

• Undergraduate background in Physics, Chemistry, and Mathematics (B.Sc. PCM).

• Trilingual expert in Hindi (native), Gujarati (fluent/bilingual), and English (professional), with working knowledge of classical Sanskrit.

By leveraging scientific epistemology and deductive logic on qualitative, historical, and multilingual data, I assist AI labs in creating high-quality datasets, stress-testing reasoning processes, and reducing hallucinations with academic rigour.

I can work properly on:

1️⃣ Ensuring historical accuracy & mitigating hallucinations: auditing model outputs against primary epigraphic, archaeological, and textual sources to remove false citations, time-period errors, and fabricated claims.

2️⃣ Chain-of-Thought reasoning evaluation: assessing logical steps in models to confirm sound deduction rather than superficial plausibility.

3️⃣ Aligning with Indic languages & cultures (Hindi, Gujarati, English): checking translation fidelity, colloquial nuances, code-switching (Hinglish/Gujlish), honorifics (Aap/Tum/Tu, Tame/Tu), and cultural safety in datasets (MILU, IndicParam).

4️⃣ Legal data training: annotating Indian legal documents (Supreme Court/High Court judgments, statutes, contracts), verifying citations (AIR, SCC), and testing for false statutes and misapplied laws.

5️⃣ Adversarial testing & bias detection: creating complex prompts to reveal subtle historiographical, regional, or Western-centric biases.

Curating datasets for ancient scripts, epigraphy, and cultural artefacts for advanced Vision-Language Models.

I use the following methods and skillsets:

• Alignment Methods: RLHF, DPO, SFT, Pairwise Model Comparison, Multi-Dimensional Rubrics.

• Core Skills: Prompt Engineering, Hallucination Detection, Error Coding (MQM), Data Curation, Legal Annotation, Paleographic Transcription, Scientific Writing.

• Domain Expertise: South Asian History, Epigraphy, Manuscriptology, Indian Law, Archaeozoology, Deductive Logic (PCM).

Curating datasets for ancient scripts, epigraphy, and cultural artefacts for advanced Vision-Language Models.

Open to impactful roles as an AI SME, Prompt Writer, Indic LLM Evaluator, Legal AI Trainer, and RLHF Specialist.

• [email]

Personal: [email]

Supervised on-site excavation trenches at the pivotal Early Historical site of Eran, managing site workers, logging structural features, recording micro-stratigraphy, and preserving delicate cultural remains.

Engaged in field investigations at Champaner-Pavagadh UNESCO World Heritage Park and completed the workshop on 'Site Presentation and Preservation', partnering with MSU Baroda, University of Allahabad, Durham University, and ASI.