§ 00 — Introduction

I build software across AI, backend systems, and infrastructure.

Software engineer with 3+ years of experience building low-latency backend services, distributed workflows, and AI-powered systems—from API and data design through deployment, observability, and production support.

CLASS
Software Engineer / AI Engineer
EDUCATION
M.S. Computer Science · Syracuse University · 2026
FOCUS
AI · Backend systems · Infrastructure
LOCATION
USA · Open to relocate
STATUS
Software Developer at Honeywell
Harsh Patel seated on a rocky overlook above a river and forested hills
FIG. 01 — THE ENGINEER IN QUESTION

§ 01

Overview

I build backend and AI systems that remain dependable under real production constraints—from high-concurrency workflows to retrieval pipelines and recommendation services.

My work spans Java and Python services, API and database design, asynchronous processing, cloud deployment, and observability. I hold an M.S. in Computer Science from Syracuse University and bring more than three years of professional software engineering experience.

§ 02

Experience

Professional experience across backend platforms, applied machine learning, distributed workflows, and production reliability.

  1. RUN-01RUNNING

    Sep 2025 — Present

    Software Developer

    Honeywell

    • Architected a seller operations platform with real-time slot enforcement, conflict detection, and role-based access control, enabling 80% self-service adoption and reducing operations requests by 45%.
    • Shipped low-latency pricing and availability services with asynchronous handlers and idempotent writes, sustaining sub-150 ms p95 response times and reducing booking conflicts by 33%.
    • Designed an LLM-powered onboarding pipeline with structured output validation, reducing seller clarification cycles by 40%.
    • Built a RAG semantic search service with embeddings and pgvector that surfaced contract guidance in under two seconds and reduced first-week support tickets by 25%.
    • Streamlined CI/CD releases with GitHub Actions, Docker, and AWS ECS using blue-green deployments, automated smoke tests, and staged rollouts, cutting deployment time from 30 minutes to under 10.
    • Improved production reliability with structured logging and automated health checks, diagnosing distributed API failures and data inconsistencies through logs, metrics, and database traces.
    • Python
    • FastAPI
    • Node.js
    • DynamoDB
    • LangChain
    • pgvector
    • AWS ECS
  2. RUN-02COMPLETE

    Jun 2025 — Aug 2025

    Machine Learning Engineer

    IConsult

    • Built a personalized clothing recommendation system using collaborative filtering and behavioral signals, achieving 78% Precision@10 with cold-start strategies for new users.
    • Developed real-time recommendation APIs and persistent interaction models for a swipe-based marketplace experience, improving click-through rate by 24%.
    • Python
    • Collaborative filtering
    • Recommendation APIs
  3. RUN-03COMPLETE

    Jan 2022 — Jul 2024

    Software Engineer

    Mindtree

    • Designed Java Spring Boot microservices and secure REST APIs for enterprise workflow management, including approval routing, task orchestration, audit tracking, and role-based access control.
    • Reduced p95 API latency from 1.8 seconds to 320 ms through query optimization, indexing, pagination, and connection-pool tuning.
    • Built asynchronous Spring Batch and Redis pipelines with idempotency, retries, and failure recovery, improving reliability by 30%.
    • Delivered a React dashboard with drag-and-drop Kanban boards and role-based views, improving enterprise client engagement by 35% and reducing onboarding time by 40%.
    • Containerized services and strengthened CI/CD workflows, reducing post-release defects by 37% across delivery cycles.
    • Led production incident resolution using log analysis, SQL diagnostics, and API tracing, then partnered with product, frontend, QA, and client teams to turn complex workflows into reusable services and well-defined API contracts.
    • Java
    • Spring Boot
    • Hibernate
    • PostgreSQL
    • Redis
    • Docker

§ 03

Projects

Engineering records with room for decisions, evidence, verified outcomes, and source links.

P-01

PaperVeil

Native macOS utility

A local macOS 13+ menu-bar utility that places a subtle, click-through paper texture across every connected display, with presets, intensity control, application exclusions, schedules, shortcuts, and URL automation.

ENGINEERING NOTE

Built the overlay to remain non-interactive and excluded from standard macOS capture paths without requiring Screen Recording or Accessibility permissions. The release workflow covers multi-display behavior, Spaces, full-screen applications, sleep/wake, signing, and notarization readiness.

  • Swift
  • AppKit
  • XcodeGen
  • App Sandbox
  • Hardened Runtime
P-02

MacTouch

Apple Silicon interaction experiment

A native macOS project that turns physical taps on an Apple Silicon MacBook chassis into configurable actions by reading the machine's internal accelerometer.

ENGINEERING NOTE

Reads AppleSPUHIDDevice reports through IOKit, decodes the accelerometer stream, groups taps into single, double, and triple gestures, and exposes calibration, thresholds, cooldowns, counters, and action mapping through a menu-bar interface.

22-byte
Hardware input reports
3
Recognized tap gestures
  • Swift
  • IOKit HID
  • Swift Package Manager
  • Apple Silicon
P-03

Credit Card Optimizer

Cardwise / financial decision system

A privacy-friendly application that turns a user's existing credit-card wallet and spending profile into category winners, backup cards, cap guidance, merchant recommendations, and an estimated annual value.

ENGINEERING NOTE

Built a deterministic reward engine over an issuer-sourced, schema-validated catalog. It normalizes annual spending, applies reward rates and caps, chooses fallbacks, separates raw points from estimated cash value, and keeps wallet data in the browser without bank credentials.

25
Supported U.S. cards
8
Supported issuers
0
Bank credentials required
  • Next.js 16
  • React 19
  • TypeScript
  • JSON Schema
  • Vercel
P-04

Geo-Climatic Prediction Pipeline

Space Applications Centre, ISRO

A machine-learning pipeline for predicting geo-climatic conditions across the Indian subcontinent from satellite, radar, LiDAR, and radiometric data.

ENGINEERING NOTE

Collected INSAT-3D satellite and DPR radar data, then normalized and flattened the heterogeneous source formats into a reusable database for model training and analysis.

  • Python
  • Scikit-learn
  • PyTorch
  • Satellite data
RESEARCH PROJECT

§ 04

Skills

Grouped capabilities without ratings, percentages, or decorative logo clouds.

SK-01

Languages

  • Java
  • Python
  • TypeScript
  • JavaScript
  • SQL
  • C++
  • Shell
SK-02

Frameworks

  • Spring Boot
  • Spring MVC
  • FastAPI
  • Node.js
  • React
  • Hibernate / JPA
  • Spring Batch
SK-03

Backend

  • REST APIs
  • Microservices
  • Distributed systems
  • Event-driven architecture
  • Asynchronous processing
  • gRPC
  • WebSockets
SK-04

AI / ML

  • LangChain
  • Retrieval-augmented generation
  • Text embeddings
  • LLM evaluation
  • AI guardrails
  • PyTorch
  • Scikit-learn
  • Hugging Face
SK-05

Cloud & Infrastructure

  • AWS Lambda
  • AWS ECS
  • API Gateway
  • SQS
  • S3
  • Docker
  • GitHub Actions
  • Jenkins
SK-06

Databases

  • PostgreSQL
  • DynamoDB
  • Redis
  • MongoDB
  • MySQL
  • pgvector
  • FAISS
SK-07

Tools

  • JUnit
  • Mockito
  • PyTest
  • OpenTelemetry
  • Prometheus
  • Grafana
  • GitHub Copilot

§ 05

Writing

Technical writing, research, case studies, and external publication records.

W-01

Engineering deep dive / Medium

MacTouch: A Native Chassis-Tap Gesture Engine on Apple Silicon

A technical walkthrough of reading Apple Silicon accelerometer reports through IOKit, processing physical chassis taps, and turning the signal into reliable single, double, and triple gestures.

W-02

System design / Medium

Building Cardwise: From an AI-Assisted Prototype to a Deterministic Credit Card Engine

How an AI-assisted prototype became a reproducible rewards engine backed by reviewed, source-controlled card data, explicit TypeScript calculations, and privacy-first local state.

§ 06

Contact

Have a system worth building? Let's talk.

Share the problem, the constraints, and what success should look like. I am based in the United States and open to relocation.