AMITH
KAMBHAMPATI.

Designed and deployed a three-workflow automation suite to manage trucking fleet compliance and financial processing, eliminating manual data entry across inspection logging, load documentation, and factoring submissions.

• Inspection Digitization Pipeline: Built a Telegram-triggered workflow that receives driver inspection photos/PDFs, converts them to images, uses Google Gemini 3 Pro vision AI to extract inspection dates and unit numbers with handwriting recognition, validates vehicle records against Airtable, and auto-updates compliance logs with annual inspection due dates.
• Rate Confirmation Processor: Engineered a web form workflow that ingests Rate Con PDFs, intelligently switches between text extraction and vision processing based on OCR quality, uses Gemini 2.5 Flash to extract load IDs, origin/destination zip codes, pickup/delivery dates, and financial breakdowns, and creates Airtable tracking records with confidence scores.
• BOL Matching & Factoring Automation: Developed an intelligent document matching system that processes incoming BOLs, extracts reference numbers and delivery locations, searches Airtable for loads awaiting BOL using load ID or zip code + date matching (±2 day tolerance), validates signatures, flags missing pages, and automatically triggers factoring submissions when document sets are complete.
• Database Architecture: Designed the Airtable schema with four interconnected tables (Fleet, Drivers, Active Loads, Compliance Log) enabling cross-referencing between vehicles, drivers, loads, and inspection records for real-time compliance dashboards.

Developed an intelligent ATS-hacking agent to automate job sourcing, screening, and application tailoring, reducing manual job search time from 6 hours to 15 minutes per batch.

• Job Aggregation Pipeline: Built a workflow using n8n and JSearch API to fetch 100+ job listings per execution with filters for remote positions, full-time roles, and experience requirements, extracting metadata including titles, companies, salaries, and application links.
• ATS Kill Switch Logic: Integrated Gemini 2.5 Pro to analyze job descriptions against a master resume, detect visa/clearance blockers ('US Citizen only', 'No sponsorship', 'Active clearance'), calculate fit scores (0-10) using a three-part rubric, and auto-reject mismatched positions before resume generation.
• Keyword Injection Engine: Engineered a system that extracts top 5 'must-have' keywords from job qualifications, identifies semantic synonym swaps, rewrites job titles and bullet points to mirror ATS language without fabricating experience, and maintains 60-70% metric density to avoid robotic formatting.
• Document Generation & Tracking: Automated Google Docs API integration to clone resume templates, populate dynamic fields via batch updates, convert to PDF, upload to Drive, append tracking rows to Google Sheets, and send batch emails with clickable application links.

Built a dual-workflow RAG system to eliminate manual knowledge management through intelligent content ingestion and conversational retrieval.

• Intelligent Content Ingestion: Engineered an n8n workflow triggered by Slack that automatically detects URLs vs. pasted text, fetches webpage content via HTTP requests, uses Google Gemini API to strip HTML noise while preserving semantic meaning in clean Markdown format, and recursively splits text into optimal chunks for embedding.
• Vector Knowledge Base: Designed a Pinecone vector database architecture with Google Gemini embeddings, enabling semantic search across saved articles, research threads, and technical documentation without manual categorization or folder structures.
• Conversational RAG Interface: Built a chat-based retrieval system using Gemini 2.5 Pro that queries the vector database with top-5 similarity search, synthesizes contextual answers from stored knowledge, and cites original sources without requiring users to remember where information was saved.
• Zero-Friction Workflow: Created an autonomous system where users drop links or text into a dedicated Slack channel, and the pipeline handles extraction, cleaning, vectorization, and indexing automatically—achieving true 'save once, query forever' knowledge management.