Aadarsh Shrestha

AadarshShrestha

Co-founder and CEO of BlackFire AI. AI engineer.

I build AI for Nepali and the low-resource languages of Nepal.

Aadarsh Shrestha in a brown jacket and navy check shirt, sitting forward in a leather armchair in front of bookshelves.

हाम्रो भाषा, हाम्रो आवाज।(Our language, our voice.)

Email meLinkedInDownload CV

My voice, cloned with Aawazz.

  1. नमस्ते, म आदर्श श्रेष्ठ हुँ।

    Hello, I'm Aadarsh Shrestha.

  2. BlackFire मा हामी नेपाली र नेपालका अन्य भाषाहरूका लागि AI बनाउँछौं।

    At BlackFire, we build AI for Nepali and the other languages of Nepal.

  3. तपाईंले अहिले सुनिरहनुभएको यो आवाज पनि Aawazz ले बनाएको हो।

    The voice you are hearing right now was made with Aawazz too.

  4. हाम्रो भाषा, हाम्रो आवाज।

    Our language, our voice.

About

I think AI can be good for people, but only when it is built into the tools they already use, in the language they speak. For Nepali, most of that work has not been done yet.

That is why I co-founded BlackFire. Before that I was a data engineer at Kantipur Media Group, then a data and AI engineer at MediaStack AI. I left in September 2026 to work on BlackFire full time, and I still build.

Aadarsh working on a laptop at a wooden table in warm lamp light.

AI, built for humans.

BlackFire (Blackfire A.I. Pvt. Ltd.) is an AI product company in Lalitpur. We build AI for Nepali and the low-resource languages of Nepal, and we start each product with the same question: what would make this useful to the person on the other side?

  • Incorporated on 22 July 2026
  • Member of NVIDIA Inception
  • Supported by AWS Activate and Cloudflare for Startups

Products

Aawazz

Live

My role: Co-creator and engineer

Nepali and English text-to-speech, voice cloning and video dubbing, on web and Android. We started it in May 2026 because AI voices kept getting Nepali wrong: conjuncts like क्ष, the vowel ऋ, and sandhi, where a word's sound changes with the words around it. We built it around Nepali phonetics from the start.

The Aawazz home page, with a demo player for Nepali voices.

Tipni टिप्नी

Public beta

My role: Came up with it and lead it

Meeting notes for Nepali, English and mixed meetings. Every note links back to the exact words it came from, and anything the transcript does not support is left out.

The Tipni home page: “Every decision has a source.”

LipiSub

Live

My role: Developer

Nepali and English subtitles for video. Transcribe, correct, edit live and export, usually in under two minutes.

The LipiSub home page, with a video shown before and after Nepali captions.

Research

  • Newari text-to-speech data

    My work at BlackFire

    A pipeline that builds one recording script for a single Newari narrator from Nepal Bhasa Wikipedia and Wiktionary data. The script is 507 lines, about 49 minutes of reading, and covers every reachable phoneme.

  • Nepali speech recognition

    BlackFire

    Two models: one that writes punctuated Nepali, and one fast enough to run on a laptop CPU.

  • A Nepali-first language model

    BlackFire, in development

    Fine-tuning open models on Nepali, English and mixed text, as a shared base for our products.

Before BlackFire

  • MediaStack AI (M&S Labs)

    Data and AI Engineer

    June to September 2026

    • Cutaway: I built the frontend for an AI tool that turns long videos into short clips with Nepali captions.
    • Newsflow: I worked across the stack on a web editor for laying out the printed newspaper, with a service that renders print-ready PDFs.
    • KMG Analytics Dashboard: I built the dashboard and analysed the data behind it.

  • Kantipur NEO, Kantipur Media Group

    Data Engineer

    April to June 2026

    My first full-time role. I consolidated a company-wide ERP into PostgreSQL and rebuilt it as a layered dbt and Airflow pipeline, so an in-house AI could use the data. Personal data was pseudonymised before the business layer.

  • eXtensoData (F1Soft Group)

    Data Engineering Intern

    August 2025 to January 2026

    Where I learned the basics: ETL pipelines and Airflow, with Spark, SQL and Linux.

Aadarsh with his M&S Labs colleagues at his farewell, a cake on the table in front of them.
My last day at M&S Labs, September 2026.

What people I worked with say

Aadarsh is thoughtful, dependable, and genuinely great to work with.

Dibbeshwor Acharya, who managed me at M&S Labs

A versatile and dependable developer with strong skills across data engineering, backend systems, APIs, and frontend development.

Ritesh Ghimire, senior colleague at Kantipur Media Group

Read them on LinkedIn

Client work

Aadarsh in graduation gown and cap at the IIMS College convocation, 2026.
Convocation, IIMS College, 2026.

Education

Bachelor of Computer Science (Hons), Taylor's College. Studied at IIMS College, Kathmandu, and graduated in 2026.

Timeline

  1. November 2024Volunteered at TEDxIIMS
  2. August 2025Started as a data engineering intern at eXtensoData
  3. April 2026Joined Kantipur NEO as a data engineer
  4. May 2026Started building a Nepali voice at night and on weekends, which became Aawazz
  5. June 2026Moved to MediaStack AI as a data and AI engineer
  6. 22 July 2026Co-founded Blackfire A.I. Pvt. Ltd.
  7. August 2026BlackFire joined NVIDIA Inception, and I demoed Aawazz at Cursor Kathmandu
  8. September 2026Left M&S Labs to work on BlackFire full time
  9. October 2026Aawazz and LipiSub live, Tipni in public beta

Speaking and community

Aadarsh speaking into a microphone on the TEDxIIMS stage, under red light.
Co-hosting the audience Q&A at TEDxIIMS, November 2024.

Writing

Posts on LinkedIn:

Contact

If you are working on something where Nepali AI could help, or you would like to invest in or partner with BlackFire, write to me.

Emailaadarsh@blackfire.com.np

© 2026 Aadarsh Shrestha. Lalitpur, Nepal.