Research Our RecurrMoE paper is published

Turn busywork into intelligent systems.

SoLS AI is an AI engineering studio. We automate operations, build production-grade software around your business, and push the frontier with our own AI research.

  • Published AI research
  • End-to-end delivery
  • Working with teams anywhere
invoice-automation.flow Live
Trigger New email · invoice_2381.pdf
09:41
AI agent Extract & validate fields Tax ID ✓$4,825.00 ✓Due 18 Oct ✓
98%
Update ERP
Notify on WhatsApp
Log to dashboard
Example automation — invoices read, validated and posted automatically

Built with the tools modern teams trust

  • Python
  • PyTorch
  • Hugging Face
  • LangChain
  • FastAPI
  • Node.js
  • React
  • Next.js
  • PostgreSQL
  • Docker
  • AWS
  • n8n
  • WhatsApp Business API
  • Flutter
What we do

Five ways we help you move faster.

From a single automated workflow to a full product team — pick where you want to start.

Automation projects

We find the repetitive work slowing your team down and replace it with reliable, AI-assisted workflows that run around the clock.

  • Document & invoice processing
  • WhatsApp, email & chat assistants
  • CRM, ERP & spreadsheet integrations
  • Automated reports & alerts
Automate a workflow

Production software

Web apps, dashboards, internal tools and SaaS products engineered to solve real business problems — and to keep running.

  • Web & mobile applications
  • APIs, dashboards & admin panels
  • Cloud deployment & monitoring
Build a product

Custom client builds

Have a spec, an idea or just a problem? We turn your requirements into working software with clear scope and weekly progress.

  • Requirement analysis & scoping
  • Prototypes & MVPs
  • Maintenance & support
Share your requirement

AI collaborations

Partner with our research team on model training, fine-tuning, applied AI and joint publications.

  • Applied AI proofs of concept
  • LLM fine-tuning & evaluation
  • Joint research & papers
How we collaborate

Internships

Hands-on programmes where students ship real code on real projects, mentored by working engineers and researchers.

  • AI/ML, full-stack & automation tracks
  • Live client & research projects
  • Mentorship & code reviews
See the programme
How we work

A clear path from idea to impact.

No black boxes. You see progress every week and know exactly what you're paying for before we write a line of code.

  1. 01

    Discover

    We learn your business, map the workflow and pinpoint where AI or software will pay off fastest.

  2. 02

    Design

    A written scope, architecture and clickable prototype — agreed with you before the build begins.

  3. 03

    Build

    Short sprints with weekly demos, tested code and progress you can actually see and try.

  4. 04

    Launch & support

    We deploy, monitor, train your team and stay on to improve the system as you grow.

Selected work

Built by SoLS, running in the real world.

RecurrMoE hierarchical architecture diagram
AI research2026

RecurrMoE — an adaptive 1B-parameter language model

Our in-house foundation model architecture that thinks harder only where it needs to, cutting compute by 44.4% at deployment.

  • PyTorch
  • Mixture-of-Experts
  • A100 training
Read the case study

Your project could be next.

Tell us what's slowing your business down. We'll show you what's possible.

Start a conversation
Research lab

We don't just use AI. We build it.

RecurrMoE is our open research into adaptive computation: a 1.03B-parameter model that decides, token by token, how much thinking each word needs — combining recursive depth with sparse Mixture-of-Experts routing. The same engineering rigour goes into every client system we ship.

Compute reduction
44.4%
Memory reduction
93.7%
Virtual expert paths
36
Pretraining tokens
12B

Bala Murugan, Deepak, Daniel Joe, Sudalai, Muthu Siva · June 2026 · DOI 10.5281/zenodo.20570332

RecurrMoE architecture: outer stage routing, inner recursive refinement and sparse expert selection
Three nested levels — stage-level early exit, adaptive recursion and top-2 sparse experts.
Why SoLS AI

Research-grade minds. Practical results.

Research-grade engineering

We designed and trained our own language model. That depth shows up in every system we build for you.

Built for production

Tests, monitoring, documentation and a clean handover — software that keeps working after launch day.

Transparent by default

Written scope, weekly demos and honest timelines. You always know where your project stands.

Partners for the long run

We stay on after launch to support, improve and scale what we built together.

Internships

Learn by shipping real things.

Our internship programme puts students and fresh graduates inside live client and research projects — not toy assignments. You'll write production code, get it reviewed, and leave with work you can show.

AI & Machine Learning Full-stack Development Automation & Integrations Data Engineering
  • Work on real client and research projects
  • One-on-one mentorship from engineers and researchers
  • Code reviews, weekly demos and career guidance
  • Certificate of completion
Apply for an internship

How to apply

  1. 1
    Send your application

    Use the contact form with your resume link, GitHub or portfolio.

  2. 2
    Short task

    A small, practical problem in your chosen track.

  3. 3
    Conversation

    A friendly technical chat with the team.

  4. 4
    Start building

    Onboard, meet your mentor and pick up your first ticket.

AI collaborations

Let's push AI forward, together.

We team up with institutions, researchers and companies who want to build or study AI seriously.

Universities & colleges

Joint research, student projects, workshops and industry exposure for your students and faculty.

Partner with us

Researchers & labs

Co-author papers, share compute and datasets, and extend our work on efficient, adaptive models.

Propose a project

Startups & companies

R&D partnerships, AI proofs of concept and co-built products where we bring the AI expertise.

Start a partnership
Our team

The people behind SoLS AI.

  • Daniel Joe

    Project Manager

  • Deepak

    Project Leader

  • Bala Murugan

    Lead Researcher

  • Sudalai

    Data Manager

  • Muthu Siva

    Data Collector

FAQ

Questions, answered.

Can't find what you need? Ask us directly.

What kind of businesses do you work with?

Anyone with a process worth improving — from local shops and manufacturers to startups and enterprises. If your team spends hours on repetitive work, or you need software that doesn't exist yet, we can help.

How does a project start?

With a short discovery call. We understand your goals, then send a written proposal with scope, timeline and cost. Work begins only once you're happy with the plan.

Do you work with clients in other countries?

Yes. We work with teams anywhere through video calls, shared workspaces and regular demos, across time zones. You are always welcome to visit our office too.

Can you work with our existing software?

Usually, yes. Most of our automation work connects tools you already use — accounting software, CRMs, spreadsheets, email and WhatsApp — rather than replacing them.

Who can apply for an internship?

Students and recent graduates in computer science, IT, electronics or related fields who are eager to build. Send us your resume and any projects through the contact form and choose "Internship".

How can my institution collaborate on AI research?

Reach out through the form and choose "AI collaboration". Tell us about your area of interest and we'll set up a call with our research team.

Contact

Let's build something that matters.

Tell us a little about what you need. We usually reply within one business day.

I'm interested in

By sending this form you agree to our privacy policy.