Looking for a job

HARSH Karumuri

@harshkarumuri

Message

Machine learning intern focused on NLP, RAG pipelines, and computer vision with production-ready engineering instincts.

India

Message

What I'm looking for

I’m looking for a role where I can build and deploy ML/NLP systems—RAG, fine-tuned LLMs, and anomaly detection—turning prototypes into reliable production workflows while growing through strong mentorship and research rigor.

I’m a B.Tech student in Electronics and Computer Engineering currently building ML systems from prototype to production. In my internships, I focused on NLP/LLMs, anomaly detection, and retrieval methods that improve real business workflows.

At Elevate Labs, I reduced manual knowledge retrieval time by 40% by building a RAG pipeline with LangChain and ChromaDB, and I improved an NLP classifier’s F1 score by 22% by fine-tuning BERT with HuggingFace Transformers on a 50K-sample domain corpus. I also increased LLM response relevance by 28% through prompt templates and context-structured data flows designed for production AI workflows.

At Jisnu Communications, I improved telemetry fault detection accuracy by 22% using a PyTorch anomaly detection model trained on 10K+ real-time sensor streams, while also speeding up training by 35% with Python ingestion and preprocessing pipelines. I complement this with projects in reinforcement learning, multimodal digital twins, and computer vision, including a DroneRL-Sim DQN agent that scaled success rate from 42% to 91% across 10K+ simulation episodes.

Experience

Work history, roles, and key accomplishments

ML Research Intern

Jisnu Communications

Jun 2025 - Sep 2025 (3 months)

Developed a PyTorch anomaly detection model for telemetry fault detection, improving accuracy by 22% over rule-based baselines using 10K+ real-time sensor signal streams. Built Python ingestion and preprocessing pipelines to accelerate training by 35%, and used NLP-based retrieval to cut research consolidation time by 30% by synthesizing findings across 30+ technical documents.

PyTorch Anomaly Detection Fault Detection Model Training Optimization Streaming Ingestion Pipelines data preprocessing Python

Machine Learning Intern

Elevate Labs

Jan 2025 - May 2025 (4 months)

Built a RAG pipeline using LangChain and ChromaDB to enable semantic search over internal documents, reducing manual knowledge retrieval time by 40%. Improved NLP classifier F1 by 22% via BERT fine-tuning with HuggingFace Transformers and increased LLM response relevance by 28% through prompt templates and context-structured data flows.