Skip to content
View TarunTomar122's full-sized avatar
🌴
On vacation
🌴
On vacation
136 followers · 63 following

Achievements

Achievement: StarstruckAchievement: Pair ExtraordinaireAchievement: YOLOAchievement: QuickdrawAchievement: Arctic Code Vault ContributorAchievement: Pull Sharkx2

Achievements

Achievement: StarstruckAchievement: Pair ExtraordinaireAchievement: YOLOAchievement: QuickdrawAchievement: Arctic Code Vault ContributorAchievement: Pull Sharkx2

Highlights

  • Pro

Organizations

@RoboticsClubIITJ

Block or report TarunTomar122

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
TarunTomar122/README.md

Hey, I'm Tarun 👋

Design Engineer at Adobe, moving to Edinburgh (Sep '26) for MSc AI.

I build on-device AI systems, VLM architectures, and design infrastructure used across Adobe products. Builder first, researcher second.

What I build

  • On-device AI — Lumi, a 350M param LLM running <100ms on Android, 400+ users on Play Store
  • Vision–language systems — NanoChat, 64× visual token compression for multimodal context extension
  • Design infrastructure — Led Spectrum Web Components migration to 2.0, impacting Firefly, Illustrator Web, and Adobe.com
  • Products — StocksBrew ($18/mo MRR), YourTrace ($28/mo MRR)

Recent work

  • Shipped privacy-first on-device LLM for Android with LoRA fine-tuning
  • Built vision–language prototype with aggressive context compression (got into target MSc program)
  • Migrated core web components used by thousands of Adobe engineers

🛠️ Stack

Frontend: React, Lit, TypeScript, Web Components, Next.js Backend: Node, Python, PostgreSQL, Supabase AI: PyTorch, LoRA, RAG, vision–language models, on-device inference

Content

  • YouTube (4.3K subs) — software engineering, AI, building products
  • Writing — tarat.space — compression, on-device AI, shipping at scale

📬 Get in touch

✨ Builder, runner (ultramarathons), always shipping.

Pinned Loading

  1. vision-pathways vision-pathways Public

    Evolutionary search and task-aware block routing for efficient vision-language models.

    Python

  2. colab-intern colab-intern Public

    Mini ML research agent: local + Colab training for models ≤1B

    Python

  3. Lumi Lumi Public

    A productivity app that feels like texting yourself

    TypeScript 38 8

  4. digital-garden digital-garden Public

    My Digital Garden

    TypeScript 7

  5. vision-encoded-nanochat vision-encoded-nanochat Public

    Forked from karpathy/nanochat

    I added a Vision Encoder to Nanochat

    Python

  6. ML-DL-implementation ML-DL-implementation Public

    Forked from RoboticsClubIITJ/ML-DL-implementation

    An implementation of ML and DL algorithms from scratch in python using nothing but numpy.

    Jupyter Notebook