# Hardik Sapra

> Senior Software Engineer & Technical Consultant specializing in high-scale backend APIs, production document AI extraction, and cloud infrastructure.

Hardik Sapra is a software engineer based in New Delhi, India. He builds high-throughput backend systems, resilient microservices, and AI-driven workflow automation tools that replace manual human data entry. He is available for select high-impact technical consulting engagements, contract engineering, and senior engineering roles.

## Core Specializations & Capabilities

- **High-Throughput Backend Architecture & API Optimization**:
  - Slashed p95 API latency by 55% at Zopper by implementing intelligent Redis multi-layer caching and offloading heavy tasks to asynchronous Celery workers and Redis Pub/Sub.
  - Architected high-concurrency microservices in Python (FastAPI, Django), Node.js, and PostgreSQL maintaining sub-200ms p95 on high-stakes transaction journeys.

- **Production Document AI & Extraction (InsureFlow AI)**:
  - Engineered an end-to-end automated intake pipeline ingesting real-world smartphone photos and multi-page PDFs over WhatsApp Webhooks.
  - Built an adaptive-resolution OCR engine using PaddleOCR (1,280px fast pass for 2.4s latency; automated 1,920px retry for dense token recovery) paired with quantized local Qwen 3.5 on CPU via llama.cpp and Claude Vision fallback.
  - Automatically extracts 20+ fields across Indian KYC (Aadhaar front/back, PAN, Passports, Driving Licenses), banking documents (Cancelled Cheques with account numbers and IFSC), and dense Motor/Health insurance schedules.

- **Deterministic Identity Resolution & Deduplication**:
  - Implemented cryptographic, document-level identity resolution (Aadhaar and PAN hashing) decoupled from transient chat sessions.
  - Built multi-parameter fuzzy phonetic matching, client deduplication, and family relationship clustering with non-destructive staging.

- **Enterprise Headless Sync & RPA**:
  - Reverse-engineered headless data sync engines with automated token refreshers for HDFC ERGO (1UP portal), Turtlemint, and Prudent Corporate Advisory (mutual fund onboarding feeds).

- **DevOps, Kubernetes & Infrastructure Automation**:
  - Reduced multi-tenant client onboarding from ~1 hour to <2 minutes with Vault-managed credentials and zero-restart activation.
  - Built and owned containerized CI/CD deployment pipelines across 7+ microservices at Twisted Mountain Animation.

## Key Case Studies

- [InsureFlow AI: Turning WhatsApp documents into organized client records](https://hardiksapra.com/work/insureflow-ai)
  - Outcome: Routine client servicing reduced from ~45 minutes to <2 minutes per document batch.
- [Backend Performance: Making insurance APIs respond faster](https://hardiksapra.com/work/api-performance)
  - Outcome: 55% lower p95 API latency under production load at Zopper.
- [Tenant Provisioning: Automated onboarding pipelines](https://hardiksapra.com/work/tenant-provisioning)
  - Outcome: Onboarding time cut from ~1 hour to <2 minutes.
- [Service CI/CD Pipelines: Full multi-service automated deployment](https://hardiksapra.com/work/service-cicd)
  - Outcome: Built and managed continuous delivery pipelines across 7+ microservices.

## Professional Experience

- **Senior Software Engineer**, Twisted Mountain Animation (Nov 2025–Present)
- **Software Development Engineer**, Zopper (Aug 2024–Jun 2025)
- **Software Development Engineer**, Instahyre (Jul 2022–Jul 2024)

## Verified Contact & Profiles

- **Website**: https://hardiksapra.com
- **LinkedIn**: https://www.linkedin.com/in/hardik-sapra
- **GitHub**: https://github.com/doctornuclear
- **Book a Call (Google Calendar)**: https://calendar.app.google/sCgvsrMBvch5bnuU9
- **Direct WhatsApp**: https://wa.me/919313972744 (+919313972744)
- **Email**: hardik.sapra1@gmail.com