RUTURAJ BHOSALE
I design and build custom LLM agents, highly accurate Retrieval-Augmented Generation (RAG) systems, and robust backend automations. Helping companies turn raw data into intelligent, self-running workflows.
Projects Directory
Production-Grade Cache-Aware RAG Engine
Developed a role-based, compliance-aware RAG pipeline featuring database-level access-control filters in ChromaDB. Implemented cache-aware prompt formatting to minimize token costs, built PyTorch Cross-Encoder neural reranking, and configured prompt injection detectors and output compliance gates.
Automated RAG Evaluation
Built a standalone Python evaluation framework to benchmark retrieval algorithms. Implemented mathematical metrics (Precision, Recall, MRR, NDCG) and model-graded checks (Faithfulness, Answer Relevance) using an LLM-judge design with automatic Gemini API failover.
Medai: AI Clinical Workspace
Developed a full-stack clinical workspace that generates structured SOAP notes, conducts dense vector scans of clinical documentation, analyzes diagnostic chest radiographs using computer vision, and runs drug-drug interaction safety checks.
Astro-Spectra (AI Spectroscopy Agent)
Designed an adaptive spectroscopic agent for astrophysicists on the Blocks.ai network. Integrates FITS data parsing, a SciPy Savitzky-Golay denoising continuum reflection loop for baseline normalization, and OpenRouter LLM chains to calculate redshift and element transitions.
The Learning Agent
Built a Chrome extension and Next.js backend that overlays learning concepts onto YouTube streams in real-time. Leverages Gemini API to extract unfamiliar terms based on a diagnostic knowledge profile and renders an always-on-top floating view using the Document Picture-in-Picture API.
Personal CRM Agent
An intelligent relationship management agent that automatically profiles contacts, parses interaction history, and generates contextual follow-up prompts. Features automatic email drafting and relationship health scoring.
Core Competencies
Generative AI Integration
Connect state-of-the-art LLMs (GPT-4, Claude), image generation models (Midjourney, Stable Diffusion), and custom APIs directly into your web and mobile applications.
RAG & Vector Search
Implement secure semantic search engines. From data parsing and chunking optimization to vector databases (Qdrant, Pinecone) for highly accurate corporate document search.
Workflow Automation & Agents
Build autonomous multi-agent pipelines using LangGraph and n8n. Automate background processes that link email, databases, custom backends, and Slack.
Technical Skills
OpenAI API
LangChain
LlamaIndex
Vector Databases
LangGraph
Start a Project
Need to build an AI product, automate key operations, or implement secure document search? Let's discuss your requirements.