AI/ML Engineer with 2 Years at Marvell Semiconductors

ROHIT RAJU KAMBLE, M.Tech – AI, IIT Patna

+91-897******* **************@****.**.** ***********@*****.*** linkedin.com/rohit-r-k-8bb19924a github.com/rohit98ckd EDUCATION

M.Tech – Artificial Intelligence, IIT Patna CGPA: 7.77 2024–26 M.Sc. – Mathematics, Karnataka University Dharwad CGPA: 8.36 2020–22 EXPERIENCE

AI/ML Intern – Marvell Technology, Aug 2025 – Feb 2026

• Built scalable LLM pipelines using GPT, Claude Sonnet, and Amazon Bedrock endpoints with prompt engineering, RLHF, and DPO to deliver context-aware workforce analytics; optimized multithreaded concurrent workflows on Databricks to parallelize prompt generation, inference, and post-processing, improving throughput and reducing end-to-end latency.

• Designed and implemented a RAG-based Talent Risk Model combining enterprise knowledge retrieval with LLM reasoning to detect attrition risk signals and workforce behavioral patterns.

• Developed survival-based attrition forecasting models (Cox Proportional Hazards, XGBoost-Cox) on 500K+ employee monthly timeline records, incorporating cohort analysis and explainable AI techniques for interpretable risk prediction. Data Scientist – One Billion Ideas, Dec 2022 – Jul 2024

• Built AI/ML applications using GPT-3/4 API integration, LangChain, and NLP-driven chatbots for database interaction.

• Implemented collaborative filtering for personalized recommendation systems. KEY PROJECTS

CUDA Kernel Fusion for Deep Learning C/C++, CUDA, Nsight Systems

• Fused MatMul, Bias Add, and ReLU into a single CUDA kernel; reduced launch overhead and improved SM utilization via shared memory tiling.

Distributed AI Workload Simulator Python, NumPy, Matplotlib

• Modeled distributed AI training (all-reduce, reduce-scatter); analyzed compute/memory/network bottlenecks across GPU/node scaling. Self-Improving RAG System Python, Transformers, FAISS, PyTorch

• Built RAG system with Hugging Face DPR and GPT-2, feedback loop via contrastive learning, and REPLUG-style hot-swappable retriever. VAE-GAN for Image Reconstruction Python, PyTorch, Pillow

• Developed a VAE-GAN model for image reconstruction on custom HEIC dataset; compared quality across latent dimensions (16–256). Clinical NLP – Information Extraction GPT-3, Few-Shot Prompting

• Few-shot system for medication, coreference, and abbreviation extraction from medical notes; outperformed supervised baselines. Mental Health App Link NLP project

• Built RNN-LSTM model using Word2Vec to identify suicidal ideation from patient text, assisting in early intervention, Project explanation video link: Mental health app. deep learning project Time Series Forecasting Python, ARIMA, SARIMA, LSTM

• Deployed ARIMA, SARIMA, and LSTM models for stock price prediction using historical data; benchmarked forecasting accuracy across approaches.

SKILLS

Languages: Python, C, C++, CUDA, SQL, Bash

ML / DL: NumPy, Pandas, Sklearn, NLTK, Spacy, Scikit-learn, PyTorch, TensorFlow, Transformers, BERT, CNN, RNN, LSTM, OpenCV GenAI & LLMs: RAG, LangChain, RLHF, DPO, LoRA/QLoRA, LLM Fine-Tuning, Agentic AI, Multi-Agent Systems GPU & Infra: CUDA, Distributed Training (DP/MP/TP), NCCL, All-Reduce, RDMA, RoCE, Docker, Kubernetes Data & DBs: PySpark, Spark, Snowflake, MySQL, PostgreSQL, MongoDB, Power BI Systems: Linux, OS Internals, Kernel, Device Drive, Multithreading,Scheduling, IPC CERTIFICATIONS & ACHIEVEMENTS

Certifications:

• AI for Medical Diagnosis – Coursera (Dec 2024)

• Distributed Systems – MIT (Sep 2025)

• LangChain for LLM Application Development – DeepLearning.AI (Dec 2024) Achievements:

• GATE DA'24 (Data Science & AI) and GATE MA'24 (Mathematics) qualifier

• LeetCode: 100+ problems solved

• Technical blog author (Semi-Supervised Learning, RL, REINFORCE)

Contact this candidate