<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Projects |</title><link>https://rupal2k.github.io/rupalportfolio/projects/</link><atom:link href="https://rupal2k.github.io/rupalportfolio/projects/index.xml" rel="self" type="application/rss+xml"/><description>Projects</description><generator>HugoBlox Kit (https://hugoblox.com)</generator><language>en-us</language><lastBuildDate>Sun, 19 May 2024 00:00:00 +0000</lastBuildDate><image><url>https://rupal2k.github.io/rupalportfolio/media/icon_hu_da05098ef60dc2e7.png</url><title>Projects</title><link>https://rupal2k.github.io/rupalportfolio/projects/</link></image><item><title>Aegis AI</title><link>https://rupal2k.github.io/rupalportfolio/projects/aegis-ai/</link><pubDate>Sat, 01 Mar 2025 00:00:00 +0000</pubDate><guid>https://rupal2k.github.io/rupalportfolio/projects/aegis-ai/</guid><description>&lt;p&gt;Aegis AI is an end-to-end group insurance underwriting platform that moves insurers from reactive, historical-claims pricing to real-time predictive intelligence.&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;The 5-step intelligence pipeline:&lt;/strong&gt;&lt;/p&gt;
&lt;ol&gt;
&lt;li&gt;&lt;strong&gt;Ingest&lt;/strong&gt; — HRMS, wearables, and clinical notes (19K+ discharge notes parsed)&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Score&lt;/strong&gt; — XGBoost engine processes 21 engineered health features with Bayesian hyperparameter optimization (Optuna, 60-trial TPE, 3-fold CV)&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Explain&lt;/strong&gt; — SHAP analysis surfaces the top 5 traceable risk drivers per employee in plain language for underwriters&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Price&lt;/strong&gt; — Asymmetric zone multipliers: up to 15% discount for healthy workforces, up to 30% loading on critical risk profiles&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Report&lt;/strong&gt; — Role-based dashboards (Underwriter Console + HR Manager Dashboard) with on-demand ReportLab PDFs&lt;/li&gt;
&lt;/ol&gt;
&lt;p&gt;&lt;strong&gt;Dual-persona workspace:&lt;/strong&gt; Underwriters see a macro portfolio of 20+ companies ranked by risk; HR managers see only their own company data, enforced via app-level RBAC.&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;Deployed on:&lt;/strong&gt; Hugging Face Spaces (Streamlit dashboard) · Render (FastAPI backend) · Neon (Serverless Postgres)&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;Security:&lt;/strong&gt; JWT bearer auth · bcrypt password hashing · HIPAA/SOC 2 aligned headers (HSTS, CSP, X-Frame-Options) · 5 req/min rate limiting&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;Stack:&lt;/strong&gt; Python · XGBoost · SHAP · FastAPI · Streamlit · PostgreSQL · ReportLab · Docker&lt;/p&gt;</description></item><item><title>Fintrak</title><link>https://rupal2k.github.io/rupalportfolio/projects/fintrak/</link><pubDate>Wed, 01 Jan 2025 00:00:00 +0000</pubDate><guid>https://rupal2k.github.io/rupalportfolio/projects/fintrak/</guid><description>&lt;p&gt;Fintrak is a personal expense tracking engine with a ₹0/month operational cost. Four n8n automation pipelines running in Docker handle the full workflow: send a receipt photo to a Telegram bot → OCR.Space extracts the text → a 12-category rule engine categorizes it → the expense logs to Google Sheets with a Google Drive receipt backup. A 9 PM IST cron sends a daily summary back to Telegram.&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;Architecture:&lt;/strong&gt; Telegram Bot (UI) · n8n on Docker (Engine) · OCR.Space API (Extraction) · Google Sheets (Database) · Google Drive (Storage)&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;Key design principles:&lt;/strong&gt; Frictionless UI (the interface lives where you already chat) · Zero monthly cost (free-tier APIs only) · Sovereign data (nothing touches a third-party server)&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;Stack:&lt;/strong&gt; n8n · Docker · Telegram Bot API · OCR.Space · Google Sheets API · Google Drive API · JavaScript&lt;/p&gt;</description></item><item><title>Local LLM Setup</title><link>https://rupal2k.github.io/rupalportfolio/projects/local-llm/</link><pubDate>Sun, 01 Dec 2024 00:00:00 +0000</pubDate><guid>https://rupal2k.github.io/rupalportfolio/projects/local-llm/</guid><description>&lt;p&gt;A local AI stack using Ollama to run open-source LLMs (Llama 3, Mistral) on personal hardware without cloud costs or data privacy concerns. Includes a Streamlit web UI that integrates the Tavily web search API for real-time information retrieval alongside local model inference.&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;Stack:&lt;/strong&gt; Ollama · Streamlit · Python · Tavily API · REST APIs&lt;/p&gt;</description></item></channel></rss>