Sachin Singh

About Sachin Singh

I’m a GenAI / AI engineering leader focused on building and scaling agentic AI solutions for regulated financial services—big banks, card issuers, and payments—turning AI initiatives into measurable SLA, capacity, and quality outcomes.I bring 18+ years in software engineering, platform delivery, and transformation (automation, Agile, DevOps), with a track record of leading global programs and teams (50+ technologists/managers across the US, India, UK, and Canada). I partner closely with Product and Operations, and align proactively with Risk, Legal/Compliance, and InfoSec to ship safely in regulated environments.Most recently, I delivered an AI customer-service workflow using agentic AI patterns (LangChain/LangSmith/LangGraph) with RAG (embeddings + vector DB) and multi-model LLM integrations (GPT, Gemini, Hugging Face). Impact: improved SLA resolution time from 2 days to 4 hours (12× faster), scaled throughput from ~100 to 300+ requests/month (and growing), and reduced SME effort per request from ~2 hours to ~10 minutes through structured review and approvals.I build production-grade GenAI with an LLMOps operating model: prompt/version control, evaluation harnesses, monitoring/observability, incident playbooks, and inference cost controls. My approach emphasizes reliability, auditability, privacy/PII controls, and governance—without slowing down delivery.I’m targeting AI Director roles in banking/card issuers where I can lead AI platform and customer-ops automation (agents, RAG, governance) that improves SLA performance, scales service capacity, and raises quality—responsibly and at enterprise scale.Human-in-the-loop (HITL) means humans review/approve/correct AI outputs in production to improve quality and reduce risk.

Experience

1. Sr Manager Software Engineering

   American Express

   Jul 2025 — Present · Phoenix, AZ, US

   Delivered an Agentic AI customer-service workflow (LangChain/LangSmith + LangGraph, RAG, GCP) that improved SLA resolution time from 2 days to 4 hours (12× faster;~92% reduction) with human-in-the-loop verification for accuracy.Increased throughput from ~100 requests/month to 300+ requests/month (+200%) post-launch, and designed the workflow to scale via automation and elastic cloud capacity as demand grows.Reduced SME effort per request from ~2 hours to ~10 minutes (12× faster;~92% reduction) by shifting SMEs to exception-based review and targeted corrections while iterating toward higher autonomy.Built and deployed the AI agent on GCP, delivered via Streamlit/Chainlit for rapid iteration, demos, and stakeholder feedback loops.Designed and implemented RAG pipelines (embeddings + vector database) and integrated multi-model LLM APIs (GPT, Gemini) plus Hugging Face transformers for task-specific workflows.Established LLMOps operating model: prompt/version control, evaluation harnesses, HITL review paths, production monitoring, incident playbooks, and cost controls for inference.​Partnered with Product, Risk, Legal/Compliance, and InfoSec to ship GenAI in regulated environments (PII handling, auditability, vendor reviews, data retention, access controls)

Education

• Visvesvaraya National Institute of Technology

  Master of Technology (M.Tech.), Heat Power Engineering

  2005 — 2007

• National Institute of Technology Kurukshetra

  Bachelor of Technology (B.Tech.), Mechanical Engineering

  2000 — 2004