Apoorv Krishna

Software Engineer

Patna, Bihar, India4 yrs 3 mos experience

Most Likely To SwitchAI Enabled

Key Highlights

Founding member of two flagship Gen AI platforms.
Achieved significant cost savings through intelligent automation.
Improved API performance by 4x using Go concurrency.

Stackforce AI infers this person is a Backend Engineer specializing in Fintech and Gen AI systems.

Contact

Skills

Core Skills

Gen AiBackend Engineering

Other Skills

API EvalsAlgorithmsAnswer EvalsArtificial Intelligence (AI)Back-End Web DevelopmentC++CeleryClaude Agent SDKConcurrent ProgrammingCronDAG architectureData ScienceData StructuresDeep LearningElasticsearch

About

- Impact-driven Backend Engineer with 3.5+ years building large-scale Gen AI and distributed systems at Razorpay, serving millions of merchants across India's leading fintech platform - Founding member of two flagship Gen AI platforms: RAY (multi-agent assistant reducing support costs by 30%) and Yogi (intelligent automation delivering ₹2.5-9.5 Crore quarterly savings by auto-resolving 100K tickets) - Deep expertise in cutting-edge Agentic AI: LangGraph, Claude Agent SDK, MCP, Graph RAG (Neo4j/Neptune), ReAct architectures, and sophisticated multi-agent orchestration combining LLM intelligence with production reliability - Proven systems engineer: Eliminated production instability, improved API performance by 4x through Go concurrency, and architected fault-tolerant distributed systems using Redis, SQS, Celery, and Kafka - Tech enthusiast obsessed with emerging technologies: Continuously exploring GenAI, cybersecurity, financial systems, machine learning, and workflow automation (N8N)

Experience

4 yrs 3 mos

Total Experience

2 yrs 1 mo

Average Tenure

3 yrs 11 mos

Current Experience

Razorpay

2 roles

Senior Software Development Engineer

Promoted

Oct 2024 – Present · 1 yr 8 mos

Yogi - AI-Powered Ticket Automation Platform
Founding team member architecting enterprise Gen AI system processing 1M quarterly tickets, auto-resolving 100K tickets and targeting ₹2.5-9.5 Crore quarterly cost savings (₹300-1000 to ₹50-100 per ticket)
Designed Master-Worker multi-agent architecture using LangGraph, Claude Agent SDK with MCP, and Claude Skills—orchestrating 4-5 specialized sub-agents with dynamic Neo4j/Neptune Knowledge Graph routing, achieving 80% accuracy on complex debugging and 98% on deterministic workflows
Architected MCP-based tool ecosystem exposing internal Razorpay APIs as standardized MCP tools, combined with Graph RAG (ChromaDB + Neo4j/Neptune) for intelligent, context-aware debugging across specialized agents
Built preprocessing pipeline and 100+ N8N workflows processing 1K-3K historical tickets per category for automated SOP generation, clustering, and human-in-the-loop validation across 20-30 ticket categories
Implemented 3-layer evaluation framework (API Evals, Answer Evals, Real-Time Self-Correction) and reduced resolution time from hours to mins.

LangGraphClaude Agent SDKMCPNeo4jNeptuneN8N+5

Software Engineer

Jul 2022 – Oct 2024 · 2 yrs 3 mos

RAY (Razorpay Assistant for You) - Gen AI Chat Assistant
Architected end-to-end Gen AI microservice using Python FastAPI, Celery, SQS, and Redis, serving 10,000+ merchants with 10K+ daily messages and achieving 30% reduction in support ticket volume
Designed sophisticated intent-based multi-agent system leveraging GPT-4 with RAG (Elasticsearch), graph-based SOP traversal using DAG architecture, evolving from V1 (10 tools) to V2 (50-60 tools) with controlled workflows
Built complete auth infrastructure (JWT session management, Redis-backed context storage, rate limiting) and implemented all support channels (chat, call, ticket) with async ticket categorization
Engineered real-time message processing architecture separating web nodes from worker nodes using -- Celery for optimized performance and scalability
Backend Engineering - Care Microservice
Eliminated production alerts from 40/day to 0/week by fixing memory leaks in Go microservices with gRPC communication and improved API performance by 4x through goroutine pool architecture
Built Request for Callback system integrating third-party providers (Concentrix, Ameyo) and enhanced -observability with custom Prometheus dashboards
Frontend Engineering
Delivered Enterprise Dashboard (React, Zustand) for 100+ clients with auth, real-time chat (Azure Bot Framework), and revamped Support Widget reducing load time by 50%, handling 25,000 daily visits

PythonFastAPICelerySQSRedisGPT-4+4

Techkriti, indian institute of technology kanpur

Hospitality Manager

Aug 2019 – Dec 2019 · 4 mos · Kanpur, Uttar Pradesh, India

Interviewed more than 150 College Students across India for the Post of Campus Ambassador.
Regularly assigned tasks and was involved in regular evaluation of students based on their performance in assignment

Analyticsschoolonline

Technical Content Engineer

May 2018 – Jun 2018 · 1 mo · Bangalore Urban, Karnataka, India

Developed unique and innovative methods for ML enthusiasts to facilitate efficient learning as
well as robust methods for assessments.
Wrote comprehensive Machine Learning articles and tutorials on Classification algorithms like Decision Trees, Random Forest Trees, Logistic Regression and on Regression algorithms like Linear Regression and Polynomial Regression.
Implemented Multi-threading Techniques in Python to reduce the pre-processing computation time from 14 Hours to 4 Hours.

Antaragni, iit kanpur

Senior Public Relations Executive

Aug 2017 – Oct 2017 · 2 mos · Kanpur, Uttar Pradesh, India

Helped in Organising the Events of National and International artists like singer-duo Vishal and Shekhar.
Took care of the Hospitality of all the Artists, their on-stage preparations.
Spearheaded a group of 1st-year students who assisted me in all activities.