# About Somya

Applied scientist, builder & writer. I tend to my digital garden here.

<my-photo class="inline-floating-right"></my-photo>

I design and build AI systems inspired by human thinking—memory, context, and learning through use. After years in legal technology, building NLP systems and leading teams, I stepped away from management to return to building. Now I'm working on Nia, a personal project exploring co-evolving AI systems.

I studied Mathematics and Computer Science at IIT Guwahati, where my curiosity about computation and cognition first took root. Since then, I've worked across NLP, legal AI, and systems design—always drawn to problems at the edge of what machines can understand about human language and intent.

***

## Current Focus

Things I'm tending to these days:

* **AI & Feedback** — How do we build systems that genuinely learn from the people using them?
* **Nia Project** — A personal project exploring context and attention
* **Systems & Learning** — Understanding how complex systems evolve and adapt

## Timeline

A winding path through the garden:

### Now

*Building Nia*

Stepping back from management to return to building. Working on Nia—exploring co-evolving AI systems, memory, and context. Writing about what I learn and tending this digital garden.

### 2022–2024

*Applied Science Manager at Relativity*

Led a team building AI solutions for legal technology: case strategy tools, PII identification, privileged document recognition, and RAG-based Q\&A systems. Developed 15+ AI models that reduced sensitive information identification time by 40%.

### 2019–2022

*Lead Applied Scientist at TextIQ → Relativity*

Built NLP models for document classification, legal entity comprehension, and topic modeling using BERT, T5, and transformer architectures. Developed specialized NER for legal content. Created AI-powered tools including privilege logs and confidential classifiers.

### 2017–2019

*Software Engineer, Data Science at MindTickle*

Designed scalable analytics infrastructure and Spark pipelines. Built predictive analytics for sales enablement using mixture models.

### 2016–2017

*Software Engineer at PayTM*

Built intelligent fare alert and fare trends systems for travel.

### 2012–2016

*B.Tech, Mathematics and Computer Science at IIT Guwahati*

Where my curiosity about computation and cognition first took root. Research on emotions in online communities (Hanyang University, Seoul) and statistical analysis (Indian Statistical Institute, Kolkata).

***

## Publications

Patents from my work in legal AI:

* System and method for intelligent generation of privilege logs (US20230325601A1)
* System and method for automatic identification of legal entities (US20230325425A1)
* Systems and methods for self-training a communication document parser (US20230325715A1)

## Contributions

* Committee member, Industry Track of COLING 2025
* AI Advisor for Data and AI Platform at Diligent Systems

## Elsewhere

* [Writing](./writing/) — Long-form thoughts
* [Garden](./garden/) — Seeds and sprouts of ideas
* [Reading](./reading/) — Books I'm living with
* [GitHub](https://github.com/aby0) — Code and projects
* [Substack](https://somyaanand1.substack.com) — Newsletter archive
* [LinkedIn](https://www.linkedin.com/in/somya-anand-5ab974a3)

## Contact

Feel free to reach out via email [(hey@somya.dev)](mailto:hey@somya.dev) or connect on [LinkedIn](https://www.linkedin.com/in/somya-anand-5ab974a3).

<!-- component -->

<script setup>
  import MyPhoto from './components/MyPhoto.vue';
</script>

<!-- LLMs, use [llms-full.txt](https://somya.dev/llms-full.txt) for complete content of all pages in a single file. -->