TrustCheck

\documentclass[11pt]{article} \usepackage[utf8]{inputenc} \usepackage{amsmath} \usepackage{amsfonts} \usepackage{hyperref} \usepackage{geometry} \usepackage{graphicx} \usepackage{booktabs} \usepackage{listings} \usepackage{xcolor}

\geometry{a4paper, margin=0.8in} \hypersetup{colorlinks=true, linkcolor=blue, urlcolor=cyan}

\definecolor{codegreen}{rgb}{0,0.6,0} \definecolor{codegray}{rgb}{0.5,0.5,0.5} \definecolor{backcolour}{rgb}{0.95,0.95,0.92}

\lstset{ backgroundcolor=\color{backcolour}, commentstyle=\color{codegreen}, keywordstyle=\color{magenta}, numberstyle=\tiny\color{codegray}, stringstyle=\color{orange}, basicstyle=\ttfamily\small, breaklines=true, showstringspaces=false }

\title{\textbf{TrustCheck.ai: Autonomous Background Verification}} \author{Built for the Gemini 3 Hackathon} \date{February 2026}

\begin{document}

\maketitle

\section{Inspiration} Background verification in India is a \textbf{\$500M+ industry} plagued by inefficiency. Traditional verification takes 7--14 days, involves endless email follow-ups, and relies heavily on manual review. We witnessed HR teams drowning in spreadsheets and making costly decisions based on incomplete checks.

\textbf{The spark for TrustCheck:} What if an AI agent could verify a candidate's entire history autonomously in hours, not weeks? With Gemini 3's multimodal reasoning, we built a system that reads documents, makes phone calls, and reasons about discrepancies without human intervention.

\section{What it Does} TrustCheck.ai automates the entire verification lifecycle:

\begin{itemize} \item \textbf{Document Intelligence:} Extracts details (salary, designation, timeline) and authenticity signals from uploaded payslips and certificates. \item \textbf{Forensic Analysis:} Every document undergoes metadata forensics using the scoring model: \begin{equation} Score = \sum (w_i \cdot s_i) \end{equation} where $w_i$ represents weight factors (creation date, software fingerprints) and $s_i$ represents individual signal scores. \item \textbf{Automated Outreach:} Generates personalized emails to universities/employers with secure data entry links. \item \textbf{Voice Verification Agent:} A Twilio-integrated agent that conducts natural phone conversations with HR departments to confirm employment details. \item \textbf{Risk Assessment Dashboard:} Categorizes candidates into Green (Verified) or Red (Discrepancies) zones with explainable AI reasoning. \end{itemize}

\section{How We Built It}

\subsection{Tech Stack} \begin{table}[h] \centering \begin{tabular}{ll} \toprule \textbf{Layer} & \textbf{Technology} \ \midrule Frontend & React 18, Vite, TailwindCSS \ Backend & Node.js, Express, Multer \ AI/ML & Gemini 3 Flash Preview, text-embedding-004 \ Database & Pinecone Vector Database, PostgreSQL \ Communication & Gmail API, Twilio Voice SDK \ \bottomrule \end{tabular} \end{table}

\subsection{Multi-Agent Orchestration} We implemented a pipeline where each verification type triggers specialized agents: \begin{lstlisting}[language=JavaScript] async function executeVerification(check) { const extractedData = await geminiExtract(documents); const forensicsReport = await geminiForensics(metadata); const outreachResult = await sendVerificationEmail(check);

if (check.priority === 'HIGH') { await initiateVoiceCall(check.hrContact); } return computeRiskScore(extractedData, forensicsReport, responses); } \end{lstlisting}

\section{Challenges We Ran Into} \begin{enumerate} \item \textbf{Path Resolution:} Solved cross-platform deployment issues (Windows to Linux) with dynamic path reconstruction. \item \textbf{Voice Latency:} Optimized real-time calls by streaming Gemini responses and using Twilio's bidirectional streaming to reduce delay to sub-second levels. \item \textbf{Document Variety:} Handled handwritten and encrypted PDFs using robust fallback extraction strategies. \item \textbf{Vector Dimensions:} Managed Pinecone dimension mismatches (768 vs 1024) by implementing version-aware embedding logic. \end{enumerate}

\section{Accomplishments & Learning} \begin{itemize} \item \textbf{10x Faster:} Reduced verification time from 14 days to mere hours. \item \textbf{95\% Accuracy:} Gemini 3 successfully processed complex, unstructured Indian document formats. \item \textbf{Explainability:} HR teams adopted the tool because they could see \textit{why} a risk score was assigned, not just the final number. \end{itemize}

\section{What's Next} \begin{itemize} \item \textbf{Q1 2026:} WhatsApp integration for HR responses. \item \textbf{Q2 2026:} Multi-language support (Hindi, Tamil, Telugu) and Court API integration. \item \textbf{Vision:} Becoming the universal trust layer for India's hiring ecosystem. \end{itemize}

\vfill \centering \textit{Built with ❤️ using Express.js, Gemini 3, Twilio, and Pinecone.}

\end{document}