CV
General Information
| Full Name | Bhushan Raju Golani |
| Location | Ann Arbor, MI, USA |
| Languages | English, Hindi |
Education
Master of Science in Data Science
University of Michigan, Ann Arbor, USA
- Relevant Coursework: Natural Language Processing, Deep Learning, Machine Learning, Database Management Systems
B.Tech & M.Tech in Mechanical Engineering (Minor in Computer Science)
Indian Institute of Technology (IIT), Kharagpur · CGPA: 8.95/10
- Relevant Coursework: Machine Learning, Computer Vision, Image Processing, Algorithms, Operating Systems
Experience
AI Research Engineer
University of Michigan · Prof. Scott Pletcher
- Architected a configuration-driven multi-agent AI platform unifying OpenAI, Anthropic, Gemini, DeepSeek, and Ollama behind a single agent interface.
- Designed a ChromaDB vector database with contextual chunking, hybrid BM25+embedding search, reciprocal-rank fusion, and cross-encoder reranking.
- Built a long-term conversation memory system with persistent ChromaDB store and semantic search over past sessions.
- Developed a React-based UI and backend endpoints for interactive multi-agent conversations.
Software Developer
Amazon · Alexa Sensitive Content Intelligence Team
- Developed scalable cloud services in Java using AWS Lambda, ECS, and DynamoDB, achieving 99.99% uptime and 50,000+ TPS.
- Prompt-engineered a vision-enabled classification service using Amazon Bedrock (Claude Sonnet 3.5) for sentiment, compliance, and risk assessments.
- Designed multimodal AI workflows integrating image retrieval and LLM reasoning with <1s latency.
- Built a ReactJS analytics app on ElasticSearch for data insights and feedback review.
Software Engineer Intern
Amazon · In-Ads-Tech Team
- Configured an Event Driven System (EventBridge) to trigger API calls for remittance file retrieval.
- Built Lambda-based microservices for file filtering and data extraction with <1s latency and 99% uptime.
Projects
Child-Centered LLM Safety Evaluation & Alignment
CSE 595 Course Project
- Designed an end-to-end LLM safety pipeline for children (ages 6–12) with a human-verified dataset based on the 4Cs framework.
- Benchmarked 10 open LLMs using GPT-5 as LLM-as-a-judge on safety, age-appropriateness, empathy, and boundary-setting.
- Trained a safety classifier and applied DPO with GPT-5 gold answers for child-centered alignment.
GPT-style Language Model & Conversational SFT
EECS 595 (NLP)
- Implemented a GPT-style transformer in PyTorch from scratch with RoPE, SwiGLU, RMSNorm, and multi-head causal self-attention.
- Built a scalable pretraining pipeline on FineWeb-Edu and SFT framework for conversational agents with selective loss masking.
e-Stethoscope: Cardiac Anomaly Detection
Master's Thesis · IIT Kharagpur
- Architected a cardiac anomaly detection system using deep learning and audio from 5,000+ stethoscopes.
- Preprocessed audio with spectrograms and MFCCs; achieved 82.08% accuracy with CNNs for heart sound classification.
Honors and Awards
Publication — Journal of Chemical Physics
Special Collection: Machine Learning Hits Molecular Simulations
- Mapping fluid structuration to flow enhancement in nanofluidic channels — physics-informed neural network models for flow enhancement prediction.
Publication — IJERT
International Journal of Engineering Research and Technology
- Prediction of Cloud Server Job Failures using ML-based KNN and LSTM — 94% accuracy for cloud job failure prediction.
Research Interests
Natural Language Processing, Large Language Models, Machine Learning, RAG Systems, Multimodal AI.
Technical Skills
Machine Learning
PyTorch, TensorFlow, Scikit-learn, Hugging Face, Transformers, LangChain, Pandas, NumPy, OpenCV
AI Technologies
Large Language Models, Vector Databases (ChromaDB), Semantic Search, RAG Systems
Programming
Python, SQL, Java, C++, JavaScript, TypeScript, CUDA
Cloud & Tools
AWS (SageMaker, Lambda, S3, ECS), Docker, Git, REST APIs, CI/CD, ElasticSearch
Other Interests
Building side projects, reading, exploring new technologies.