Hello! đź‘‹

I'm Vedant Singh

Software Engineer

About Me

Hey there! I'm Vedant, a computer science student at IIIT Vadodara who loves building products from the ground up that people actually use. Whether it's crafting smooth mobile experiences with Flutter and SwiftUI, building real-time tracking pipelines, designing web interfaces with React, or diving into the mathematical beauty of LLMs and AI integrations—I'm all in. I believe the best code is the one that solves real problems and makes life easier. When I'm not turning coffee into code, you'll find me exploring new tech, contributing to open-source, or experimenting with novel architectures.

Flutter SwiftUI TypeScript React.js Next.js Three.js Node.js TensorFlow LangChain LangGraph RAG LoRA/PEFT Firebase MongoDB MySQL Pinecone Git Java

Education

Bachelor of Technology in Computer Science and Engineering

Indian Institute of Information Technology (IIIT), Vadodara

Nov 2022 - May 2026
  • Current CGPA: 8.97
  • Involved in research on topological data analysis, LLMs, and mobile systems design
  • Active in open-source and cross-platform mobile development initiatives

Minor Equivalent - Artificial Intelligence

IIT Madras (NPTEL)

Nov 2023 - May 2026
  • Completed advanced coursework in AI, ML, and Data Science under IIT Madras faculty
  • Applied learned concepts in real-world research and development projects
  • Expanded academic foundation in neural networks and intelligent systems

Experience

LLM Research Intern

Indian Institute of Information Technology (IIIT), Vadodara

Nov 2025 - Mar 2026
  • Designed a 12M-parameter encoder-decoder Transformer to generate natural language activity descriptions from WiFi CSI signals, achieving a BLEU-4 of 0.67 (~50% over baseline).
  • Engineered multi-scale temporal convolutions, Rotary Position Embeddings (RoPE), and stochastic depth on the UT-HAR dataset, reaching 96.2% activity accuracy.
  • Implemented curriculum learning and CSI-aware data augmentation (time warping, channel dropout), improving convergence by +0.08 BLEU-4 over non-curriculum training.

Mobile SDE Intern

Merlin AI (Google-backed)

May 2025 - Oct 2025
  • Built image generation flows for Merlin AI (Flutter), powering Bonkers' mobile experience; optimized prompt parsing and rendering pipelines, reducing latency by ~26%.
  • Built core modules for Thine in SwiftUI, a privacy-first AI co-pilot; engineered its task router for seamless switching between chat, image gen, and doc tools.
  • Engineered immersive native widgets using Jetpack Glance (Android) and WidgetKit (iOS) via method channels, overcoming Flutter's limitations and increasing daily active widget interactions by 2.3x.

Research Intern

Indian Institute of Information Technology (IIIT), Vadodara

May 2025 - Aug 2025
  • Optimized persistent homology for large-scale topological data analysis, reducing runtime by ~30%.
  • Utilized Ripser to compute Vietoris–Rips complexes on 68K+ point-cloud data, streamlining homology group extraction by addressing computational bottlenecks.
  • Implemented parallel matrix reduction and domain decomposition strategies, scaling computation to 2.5Ă— larger datasets while reducing runtime by 30% on complex point-cloud data.

Software Developer Intern

Zyra

Jan 2025 - Apr 2025
  • Built real-time order tracking and push-notification-driven delivery workflows in the fulfillment app, handling 10K+ daily orders with sub-second state sync across driver and customer apps.
  • Developed an AR-based virtual try-on feature using ARCore and on-device pose estimation, letting users visualize clothing fit in real time with <100ms overlay latency and 30fps body-mesh tracking.
  • Built a server-driven UI framework for dynamic catalog and promotion rendering across 50K+ SKUs.