Devansh Singh

About Devansh Singh

I’m passionate about building real-world AI systems that make technology smarter and easier to use.At Accenture, I design and develop Generative AI solutions using Large Language Models (LLMs), LangChain, and Retrieval Augmented Generation (RAG). My goal is to turn complex ideas into practical tools that improve accuracy, speed, and decision-making.I began my journey in technology at VIT Vellore, where I discovered my love for artificial intelligence. Since then, I’ve focused on Natural Language Processing (NLP) and automation from data extraction to intelligent QA systems.One of my proudest moments was when I built AI-based detectors that improved validation accuracy by over 40%. It showed me how data-driven innovation can make a measurable impact on performance and quality.I enjoy solving challenging problems, learning new tools, and collaborating with people who share a passion for technology.If you’d like to connect or discuss ideas in AI, feel free to reach out! Email: d••••••••@gmail.com Portfolio: https://ai-dev-devanshs288ds.replit.app/Skills:Large Language Models (LLMs) · LangChain · Retrieval-Augmented Generation (RAG) · Natural Language Processing (NLP) · AI-driven Automation · Python · Streamlit · FastAPI · SQL · Prompt Engineering · HTML/PDF Extraction · Document Understanding · Multi-Agent Systems · Generative AI Applications

Experience

1. Artificial Intelligence Engineer

   Accenture

   Sep 2023 — Present · Gurugram, IN

   ApprovalX – EmailQA Channel (Marketing Email Validation Platform)• Built and optimized an AI-powered detector framework and a Playwright-based HTML→PDF pipeline with coordinate-level mapping, improving validation accuracy from 40% to over 80% and enabling pixel-precise semantic comparisons between reference PDFs and marketing emails.• Implemented multilingual QA support for French, German, and Spanish using FastText + langid, reducing false positives by 72% and delivering production-ready outputs for GSK and Merck while collaborating closely during UATs to enhance reliability. Unified Marketing Platform (LangGraph-based Agentic AI QA Automation)• Architected and scaled a LangGraph-based 4-agent system (Ingestion, Branding, Functional, Compliance) that unified EmailQA, BannerQA, and MLR channels under a modular, cross-channel QA automation platform for BP and JLR• Modularized detector logic for parallel QA execution, cutting latency by 40%, and mentored junior engineers on LangGraph workflows, reducing onboarding time by 35% and strengthening the internal AI team’s expertise.

Education

• Vellore Institute of Technology

  BTech - Bachelor of Technology

  2019 — 2023