Software Engineer
Durgesh Tiwari

Software Engineer

Building high-performance systems with Swift, C++, and Metal. Focused on GPU-centric rendering, low-latency architecture, and experiences that feel effortless.

Scroll
About

I build systems that feel effortless.

I'm a Software Engineer at Kahana, where I work on GPU-centric rendering pipelines, multi-threaded architectures, and AI-assisted interactions for spatial computing. My work spans Swift, C++, Metal, and real-time systems targeting 90 FPS on visionOS.

I hold a Master's in Data Science from Indiana University Bloomington and a B.E. in Information Technology from the University of Mumbai. I believe great software is invisible — it just works.

0
Years Experience
iOS, C++, systems engineering
0
GitHub Projects
Open source contributions
0
FPS Target
GPU-centric rendering
0
Search Speedup
Trie + HashMap engine
Projects

Digital Craftsmanship

A curated collection of projects that demonstrate my passion for building elegant, high-performance software.

Software Engineer — Kahana

Kahana Vision Browser

A high-performance spatial interaction engine for visionOS. Built a GPU-centric, multi-threaded rendering pipeline achieving sustained 90+ FPS with Metal, processing ML inference asynchronously while maintaining smooth frame pacing.

SwiftC++MetalObj-C++ONNX RuntimeSwiftUI
//

90 FPS Rendering

GPU-centric pipeline with Metal command buffers and tile-based rendering

//

Multi-Threaded Architecture

Lock-free input, asset, inference, and render threads with snapshot model

//

AI Voice Assistant

Command processing with NWPathMonitor for network resilience

//

35% Frame Time Reduction

Moved scroll math to vertex shader, reduced CPU uniform updates

Software Engineer — Programmers Army

High-Performance Search Engine

I wanted to create something that wasn't just fast, but also smart enough to handle typos. So, I implemented a Trie data structure combined with the Levenshtein Distance algorithm (for fuzzy matching) in C++. To make it usable on the web, I wrapped the C++ core in a Node.js addon using node-addon-api. It's designed to be a raw, developer-focused tool—no fancy marketing fluff, just performance metrics and results.

C++STLMultithreadingNode.jsGDBValgrind
//

Trie + HashMap Architecture

Prefix matching with constant-time cached lookups

//

Fuzzy Matching

Levenshtein distance for typo tolerance up to 4 characters

//

Thread-Safe Concurrency

Mutex locks, condition variables, shared_ptr for parallel reads

//

Automated Memory Audits

GDB and Valgrind workflows for runtime memory inspection

Systems Project

Low-Latency Trading Simulator

An event-driven trading simulator in C++ on Linux with lock-free ring buffers, in-memory order book with price-time priority matching, and microsecond-level latency optimization.

C++LinuxAtomicsperfValgrindGDB
//

Lock-Free Architecture

Ring buffers with atomic operations, zero mutex contention

//

Order Book Engine

Price-time priority matching with deterministic execution

//

Cache-Optimized Hot Paths

Pre-allocated buffers, cache-line aligned data layout

//

Sub-11ms Frame Budget

Profiled with Linux perf for CPU hotspots and branch mispredictions

Computer Vision Project

Motion Estimation Pipeline

Frame-to-frame motion estimation using block matching with SAD, CUDA-accelerated GPU kernels, and exponential temporal smoothing for stable real-time optical flow.

C++CUDAImage ProcessingGPU Computing
//

Block Matching + SAD

16x16 block search with Sum of Absolute Differences

//

CUDA Acceleration

One GPU thread per block for parallel motion vector computation

//

Temporal Filtering

EMA smoothing with threshold-based gating for noise rejection

//

Real-Time Optical Flow

Stable motion field output for camera and object tracking

iOS Application

Catalog App

An iOS catalog app integrating UIKit components into SwiftUI via UIViewRepresentable, with CoreData for offline-first persistence of products, favorites, and cart state across sessions.

SwiftUIUIKitCoreDataCombineMVVM
//

UIKit + SwiftUI Bridge

Custom UIViewRepresentable for dynamic tiles and navigation

//

CoreData Persistence

Offline-first storage syncing favorites and cart across launches

//

Reusable PersistenceManager

Centralized data layer for consistent access patterns

//

Tab Navigation + Search

Dynamic filters and tab-based browsing experience

Skills

Technologies I Work With

From low-level systems programming to modern UI frameworks — tools I use to build performant, reliable software.

Swift
SwiftUI
UIKit
C++
Metal
Python
Combine
CoreData
Swift
SwiftUI
UIKit
C++
Metal
Python
Combine
CoreData
MVVM
ARKit
RealityKit
PostgreSQL
MongoDB
Docker
AWS
Kubernetes
MVVM
ARKit
RealityKit
PostgreSQL
MongoDB
Docker
AWS
Kubernetes

iOS & Mobile

SwiftUI, UIKit, Combine, CoreData, AVFoundation, CloudKit, ARKit, RealityKit, Auto Layout

Systems & Performance

C++ (14/17/20), STL, Metal, CUDA, Multithreading, Smart Pointers, GDB, Valgrind

Infrastructure & Tools

Docker, Kubernetes, AWS (S3, Lambda, CloudFront), CI/CD, Git, Linux, RabbitMQ

Experience

My Journey

From Mumbai to Bloomington to Chicago — building systems that prioritize performance and user experience.

Jun 2025 — Present

Software Engineer

Kahana, Chicago, USA

Building a GPU-centric rendering pipeline for visionOS achieving 90+ FPS. Developing SwiftUI + Combine reactive views with MVVM, integrating AVFoundation with NWPathMonitor for resilient voice commands, and implementing secure session management with Keychain and CloudKit.

SwiftC++MetalSwiftUIvisionOS
Aug 2024 — May 2025

Head Teaching Assistant

Indiana University Bloomington, USA

Mentored 100+ graduate students in iOS development, guiding them through SwiftUI, UIKit, and MVVM architecture. Reduced student error rates by 15% through hands-on debugging and troubleshooting support.

iOSTeachingSwiftUIUIKit
Mar 2022 — Aug 2023

Software Engineer

Programmers Army, Mumbai, India

Built a Trie and HashMap-based search engine reducing search time by 40% and memory usage by 30%. Developed multi-threaded programs with shared_ptr, mutex locks, and condition variables. Automated GDB and Valgrind workflows improving throughput by 20%.

C++Node.jsMultithreadingLinux
Aug 2023 — May 2025

M.S. in Data Science

Indiana University Bloomington

GPA: 3.6/4.0. Coursework in Applied Algorithms, Software Engineering, Cloud Computing, Applied Machine Learning, and Computer Vision.

AlgorithmsMLCloudComputer Vision
Aug 2018 — May 2022

B.E. in Information Technology

University of Mumbai, India

GPA: 9.3/10.0. Coursework in Object-Oriented Programming, Computer Networks, Operating Systems, and Statistics.

OOPNetworksOSStatistics
Contact

Let's Build Something Together

Have a project in mind or just want to chat? I'm always open to new opportunities and interesting conversations.

+1 999 999-8989