Skip to content
View devesrawat's full-sized avatar

Block or report devesrawat

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
devesrawat/README.md
Devesh Rawat — Production Agentic AI for Regulated FinTech

Senior Software Engineer building production agentic AI for regulated fintech — banking payment rails, compliance-aware RAG, and LLM systems built to survive an audit, not just a demo.

GCC · UK · India · Toptal Top 3% · Remote-first (IST, UTC+5:30)

Portfolio LinkedIn The Failure Mode Email


Animated signal: detected, recovered, stable — the failure-mode motif

// production incident, live system  ·  detected → recovered → stable


What I ship

Eight years shipping distributed systems; now building LLM-powered agents for banks, fintechs, and IoT fleets across GCC, UK, and India. Recent war stories:

  • Building the telematics AI agent at Nexular — LangGraph multi-agent orchestration + hybrid RAG over high-volume CAN-bus IoT telemetry from electric two/three-wheeler fleets across India, Africa & Southeast Asia, with Redis semantic caching and complexity-based multi-model routing.
  • Cut LLM inference cost with semantic caching — a two-layer cache (exact + semantic) stopped the majority of repeat-intent queries from ever reaching the model, without sacrificing correctness in a regulated support context.
  • Shipped BACS Direct Credit/Debit payment infrastructure integrating ClearBank, Temporal Cloud, and Mambu core banking for a GCC fintech — eliminating legacy workflow failures across production financial products.
  • Took a Rust media pipeline from 800ms → under 200ms while cutting infra cost 25%, and pushed webhook delivery reliability up 95% with an SQS + DLQ redesign.
  • Delivered a Sharia-compliant Instant Savings Account end-to-end for a GCC banking product, alongside compliant multi-currency conversion via Currency Cloud.

Stack

Regulated FinTech & Payments
BACS ClearBank Mambu Currency Cloud Temporal

AI / Agentic
LangGraph CrewAI Anthropic SDK MCP RAG Groq

IoT & Telematics
IoT Telematics CAN Bus Fleet Management Hybrid RAG

Languages & Infra
Rust TypeScript Python PostgreSQL AWS Kubernetes


Engagements

🏗️  Production Builds 🔍  Technical & Compliance Audits ⚙️  Fractional Architecture
End-to-end delivery of agentic AI and payment infrastructure — architecture through production. Architecture reviews for teams shipping LLM systems into regulated environments. Ongoing technical ownership for teams that need a senior architect on tap.

Writing the war stories at The Failure Mode — production agentic AI, one failure mode at a time.

Pinned Loading

  1. refinableco/refinable-sdk refinableco/refinable-sdk Public

    A drop-in SDK to interact with NFTs on the Refinable platform. Mint, buy, sell NFTs programatically.

    TypeScript 5 1

  2. MCPForge MCPForge Public template

    Rust 1

  3. dsa-tracker dsa-tracker Public

    HTML

  4. my-portfolio my-portfolio Public

    HTML

  5. payments payments Public

    JavaScript