Chinmay – Workflow Automation, SQL, n8n

An AI-powered lead qualification and response orchestration platform designed to help sales teams process inbound inquiries faster while maintaining compliance and human oversight. The system uses a multi-agent workflow to validate, triage, score, draft, review, and route incoming messages across multiple communication channels, automatically handling routine requests and escalating high-value or high-risk cases for human approval. It provides real-time operational visibility, auditability, team notifications, and resilient fallback mechanisms to ensure uninterrupted service.

• Architected a six-agent LangGraph pipeline for intake validation, triage, lead scoring, reply drafting, compliance screening, and routing;
• Designed human-in-the-loop governance, requiring approval for high-value and high-risk items before replies were sent;
• Integrated four communication channels, including a two-way Telegram bot, web lead widget, transactional email, and n8n/Zapier webhook workflows;
• Built real-time observability with SSE-streamed reasoning traces, an operations dashboard, a full audit trail, and Slack hot-lead alerts;
• Engineered a deterministic fallback engine to ensure zero downtime during LLM provider failures;
• Wrote comprehensive pytest API test coverage and deployed the platform to production using Render Blueprint infrastructure as code;
• Achieved an average pipeline processing time of approximately three seconds from message arrival to a compliance-checked draft.

An AI-powered investment research platform that transforms natural-language investment questions into structured, evidence-based analysis for retail investors. The system combines real-time market data, news intelligence, transparent scoring models, and user risk profiles to generate investment briefs with clear recommendations, expected return scenarios, supporting evidence, and key risk factors. It provides source-backed insights and personalized stock recommendations to help users make more informed investment decisions.

• Architected a nine-node LangGraph supervisor pipeline for routing, intake validation, query framing, evidence gathering, market data processing, scoring, drafting, validation, and retry handling;
• Implemented dual-source market data retrieval with graceful degradation using a primary provider and fallback quote API to ensure uninterrupted service;
• Built evidence-grounded analysis generation, with source citations, validation checks, and automatic retry mechanisms to enforce output completeness;
• Designed a transparent scoring engine that explains how market data, news signals, price movements, and user risk profiles contribute to final recommendations;
• Developed the React frontend with profile-gated workflows, personalized stock recommendations, structured analysis views, and live agent debugging traces;
• Deployed the platform to production on Render using separate FastAPI and static frontend services with environment-based configuration and CORS management;
• Hardened LLM output parsing through resilient JSON extraction and implemented ticker resolution from natural-language company names.

An AI-powered lead generation and communication automation platform designed to help businesses capture, qualify, and engage prospects across multiple channels. The system consolidates inbound leads from social platforms and web forms, applies real-time AI-based scoring, and automates personalized outreach, lead routing, and meeting scheduling. By centralizing communications and reducing manual follow-up, it enables faster response times, improved conversion rates, and greater operational efficiency.

• Built a fully automated n8n workflow that captured and scored leads from LinkedIn, Instagram, Twitter/X, and web forms in real time using the Claude API;
• Reduced lead response time from 4–24 hours to under 60 seconds through AI-powered automated replies and intelligent lead routing;
• Unified incoming communications from Outlook, SMS, Slack, and social direct messages into a single dashboard, reducing daily manual monitoring efforts;
• Automated meeting scheduling by integrating Calendly webhooks into the hot-lead workflow, eliminating manual booking tasks;
• Logged all lead interactions to Google Sheets to provide complete pipeline visibility and reporting capabilities;
• Delivered a threefold increase in lead-to-meeting conversion rates while ensuring no leads were missed across connected channels.

An AI-assisted fraud detection platform for the banking and lending sector that analyzes high volumes of financial transactions to identify potentially fraudulent activity in near real time. The system combines rule-based detection with AI-powered risk assessment to prioritize suspicious transactions, generate investigation cases, and provide explainable insights for fraud analysts. It helps financial institutions accelerate fraud investigations, improve detection accuracy, and reduce operational workload.

• Designed and deployed a hybrid fraud detection system combining a SQL Server rule engine with Azure OpenAI to monitor approximately 85,000 daily banking and lending transactions in real time;
• Integrated Azure OpenAI for advanced pattern analysis on rule-flagged transactions, generating risk scores and plain-language explanations for fraud alerts;
• Reduced the false positive rate from 18% to 10.5% by layering AI-driven classification on top of rule-based detection outputs;
• Cut mean time to detect fraud from 48 hours to under 2 hours through a real-time automated processing pipeline built in Python;
• Automated investigation case creation within 30 seconds of fraud detection, replacing a previously manual case generation process;
• Built a real-time Power BI dashboard to track flagged transactions, risk scores, and investigation status across the fraud workflow;
• Identified and escalated approximately $2.3 million in potentially fraudulent transactions during the first quarter after deployment.