Open to opportunities

Sharvary H H

@sharvaryhh

Message

AI/ML engineer building production LLM pipelines, multimodal medical AI, and scalable backend APIs.

India

Message

What I'm looking for

I’m looking for a team where I can ship production AI systems—LLM/RAG pipelines, multimodal models, and robust APIs—with strong engineering practices like CI/CD and automated testing.

I built a production-grade sub-vertical tagging system at Draup Pvt. Ltd., using LLM-based classification plus semantic similarity to retag and canonicalize a 432-class taxonomy. I found a 26% label-error rate in research-tagged ground truth, audited it with LLM assistance, and improved name-handling by diagnosing company-name hallucinations and adding name masking at finetune + inference.

At Bock Bharath, I worked on EIRA 2.0, a production-grade multimodal medical AI pipeline that combines BLIP image captioning with LLaMA grounded response generation. I also fine-tuned LLaMA on domain-specific medical PDFs using instruction-tuning to better handle unstructured document data.

Alongside internships, I’ve shipped full-stack and testing work like AutoTestKit, an automated API testing framework with pytest and JIRA REST API logging wired into GitHub Actions, plus TabMind, an installable Chrome extension for AI-powered semantic tab clustering.

Experience

Work history, roles, and key accomplishments

ML Intern

Draup Pvt. Ltd.

Feb 2026 - Jun 2026 (4 months)

Developed a production-grade sub-vertical tagging system for large unstructured datasets using LLM-based classification and semantic similarity. Found a 26% label-error rate in research-tagged ground truth via LLM-assisted audit and retagged a 432-class taxonomy.

LLM Classification EDA GTE Large Python

AI Intern

Bock Bharath

Nov 2024 - Feb 2025 (3 months)

Built EIRA 2.0, a production-grade multimodal medical AI pipeline using BLIP for image captioning and LLaMA for grounded response generation. Fine-tuned a LLaMA model on domain-specific medical PDFs using instruction-tuning.

BLIP LLaMA Deep Learning Instruction Tuning Python