Headline

Software Engineer @ AT&T | Applied AI & Python Backend | Agentic AI, LangGraph, FastAPI, RAG | 6+ yrs

About

I build production software that takes expensive manual work off people's plates, and more and more of that work now runs on AI agents. At AT&T, I work on an agentic AI platform that automates internal workflows. My focus is the Python backend (agent execution, state management, and orchestration in LangGraph, on GPT and Claude models) and the retrieval, evaluation, and guardrail layers that keep it reliable. Before AT&T, I spent five years at Verizon building B2B pricing, deal approval, and document automation tools: • A parallel OCR tool that pulls contract metadata from hundreds of legal documents per run, sending only low-confidence results to human review • Pricing model generation cut from 15 seconds to 3 seconds, and data retrieval time by up to 70% • A promo platform, grown from my team's hackathon idea (3rd of 150 teams, $1M in internal funding), that cut best-case approvals from 4 weeks to 7 days • A platform unifying 7 data sources that saves 100+ hours a month across 100+ users • A text-to-SQL workflow and a RAG pipeline on GCP Vertex AI I also led and mentored a team of four Python developers. I work mostly with Python, FastAPI, SQL, LangGraph, LangChain, RAG, Vertex AI, Tesseract OCR, Docker, and Jenkins. I'm most interested in problems where AI has to work reliably in production, not just in a demo. Email: anurajsahill99@gmail.com

Location

Bengaluru

AI stack

codex

github copilot

hugging face

langchain

langfuse

langgraph

Work experience

Software Engineer at AT&T, Aug 2025 - Present · 1 yr 2 mos, Bengaluru, Working on an agentic AI platform that automates and manages internal workflows for AT&T employees. • Build the Python backend for agent execution, state management, and orchestration, using LangGraph with GPT models • Work on retrieval, output evaluation, guardrails, and retry and fallback handling for multi-step agent.

Software Development Engineer - II at Verizon, Jul 2022 - Aug 2025 · 3 yrs 2 mos, Hybrid, Built B2B pricing, deal approval, and document automation platforms, and led a team of 4 Python developers. • Built a parallel Tesseract OCR tool that extracts contract metadata (amendments, acknowledgement letters, entity agreements, legal notices) from hundreds of contract and handwritten documents per run, with confidence scoring that routes only uncertain extractions to human review • Cut pricing model generation from 15 s to 3 s and data retrieval time by up to 70%, optimized the Python APIs, and set up infrastructure on OCI 2.0 • Turned our hackathon idea (3rd of 150 teams, $1M in internal funding) into a production promo platform, cutting manual promo creation from a week to a day and best-case approvals from 4 weeks to 7 days • Led development of a platform consolidating 7 data sources for multi-million-dollar deals, saving 100+ hours a month across 100+ users • Built an approval workflow engine for $1M+ deals across Finance, Pricing, and Contracting teams • Built a text-to-SQL workflow with Streamlit and a RAG pipeline on GCP Vertex AI • Owned code reviews, coding standards, and onboarding docs that had new engineers shipping code in their first week

Software Development Engineer - I at Verizon, Aug 2020 - Jul 2022 · 2 yrs, Hyderabad, • Built margin calculation APIs and Python pricing calculations for a customer pricing portal used in executive deal reviews • Automated executive review document generation for $1M+ deals, cutting processing time by 90% • Automated complex manual Excel work for business users, saving them hundreds of hours a month, and extended the open-source xlcalculator library into an in-house package • Restructured the codebase, halving deployment and validation time • Worked across Spring Boot, Oracle DB, and React as needed

Student Intern at Verizon, Feb 2020 - Aug 2020 · 7 mos, Hyderabad, During my internship at Verizon in Hyderabad, India, I had the opportunity to work on Springboot microservices, Angular JS, and Oracle DB. I collaborated with various stakeholders to understand the company's operations and developed rest responses for pricing products.

Public links

github: https://github.com/randnumX

medium: https://medium.com/@anurajsahill99

youtube.com: https://youtube.com/@randnumx