Hariom Vyas

About

Forward-deployed, full-stack/platform engineer who owns production features end-to-end across frontend, backend, data, and infrastructure (AWS serverless, Terraform, React). Has modernized critical fintech/payment systems (zero-downtime monolith-to-microservices with Kafka event sourcing) and productionized AI-native support workflows (LLM + RAG on Pinecone) with measurable gains in latency, incidents, CSAT, and support efficiency.

Key Strengths

• Owned and shipped self-service analytics end-to-end (React + AWS serverless + IaC), reducing support tickets 40% and cutting report time from ~2 minutes to <10 seconds
• Strong performance engineering and reliability practices (integration/load tests, CloudWatch/Datadog monitoring, alarms, runbooks)
• Resolved complex data/latency issues by adding ETL with AWS Glue to Parquet and optimizing Lambda cold starts with provisioned concurrency + Redis (60% avg latency reduction)
• Productionized LLM/RAG support assistant with Pinecone and guardrails; achieved 35% faster ticket resolution and +20% CSAT; maintained >90% weekly eval accuracy
• Built operational mechanisms to prevent stale AI answers (doc-version tagging + automated retraining triggers), reducing stale responses 70%
• Accelerated large refactors using AI-assisted dev tools with rigorous quality gates (lint/type checks, SonarQube, CI coverage/complexity thresholds), cutting timeline from 3 weeks to 5 days and reducing runtime errors 40%
• Led zero-downtime monolith-to-microservices migration using strangler pattern, shadow traffic, and regression testing; 45% incident reduction post-migration
• Handled data consistency during partial cutovers via Kafka-based event sourcing; improved P99 latency 35% with circuit breakers and caching
• Partnered with CPO on ambiguous usage-based billing roadmap; reduced billing disputes 60% and improved NPS by 15 points
• Effective forward-deployed execution across two startups with minimal oversight; shipped MVPs in 6 weeks and reduced coordination overhead ~30% via async-first tooling
• Designed and deployed Kafka + Spark Streaming pipeline processing millions of events/day
• Zero-data-loss migration using blue-green deployment, replication/acks, idempotent producers, and Spark checkpointing
• Improved streaming processing latency by 60%+
• Cross-layer debugging across software, on-prem hardware, networking, and cloud endpoints (identified faulty switch port)
• Built customer-configurable Python validation framework via YAML (no-code rule configuration for non-technical users)
• Performance optimization with multiprocessing for large datasets
• On-site customer/operator collaboration with short feedback cycles, daily standups, and rapid prototyping
• Reduced warehouse sensor-to-database data latency by 70%+ using MQTT and monitoring dashboards
• Strong in cloud + IoT integrations connecting PLC/MES to AWS/Azure for real-time analytics

Experience

Education

Skills

Languages