Government● Live
Department of Museums · Tamil Nadu
Live AI and immersive technology delivered across five government museums — Chennai, Kanchipuram, Virudhunagar, Nagapattinam, Madurai Gandhi Museum.
Marxen · Chennai · Est. 2024
The AI you own.
Not the AI you rent.
Marxen is an AI deployment company — we design, deploy, and operationalise sovereign AI inside Indian enterprises and government institutions. We do not resell. We do not just research. We ship working systems your team can run.
- Designthe architecture
- Deployon your hardware
- Operateuntil handover
12+ industries deployed5 government museums delivered1 airport authority — ChennaiLive B2B in healthcare · law · educationPan-India · Nepal · BangladeshIndic-first by engineeringDPDP-aware from designAir-gapped where required
§ 01The premise
Most AI built for India was not built in India.
It was adapted from foreign platforms, hosted on foreign servers, and priced in foreign currency. It sends payroll records, patient reports, legal filings, and citizen data outside the country — often without anyone deciding that on purpose.
Marxen was built on a different premise. We design AI systems from first principles for the Indian context — data-resident, language-aware, regulation-ready, engineered to run on infrastructure you control. Not translated. Not white-labelled. Not leased back to you as a subscription.
We are a deep-tech company. We build the model layer, the data layer, and the application layer — and we hand the whole thing to your team.
§ 02What we do
Four pillars. One sovereign standard.
A vendor sells you access. A partner builds you capability — and then makes sure you can run it without them.
§ 03The stack we operate
If your data matters, the stack matters too.
Sovereign-AI is an opinion about infrastructure. These are the open-source and Indic-first tools Marxen builds with — every deployment is auditable down to the model weights, the vector store, and the container image.
Modalities ·SpeechTextDocument · OCRImage · VisionVector retrieval
- vLLM
- Llama
- Mistral
- Hugging Face
- Gemma
- Qwen
- DeepSeek
- Phi
- Falcon
- OLMo
- Granite
- Yi
- Command R
- Cohere
- Anthropic
- Claude
- OpenAI
- Stability AI
- Replicate
- Together AI
- Fireworks
- Groq
- Perplexity
- RunwayML
- Pixtral
- LLaVA
- Florence-2
- PaliGemma
- BLIP-2
- SmolLM
- Mixtral
- Zephyr
- Nous Hermes
- Vicuna
- Aya
- Solar
- ChatGLM
- Whisper
- Distil-Whisper
- MMS
- Sarvam AI
- AI4Bharat
- Bhashini
- IndicTrans2
- Krutrim
- BharatGPT
- OpenHathi
- Project Vaani
- IndicBERT
- IndicWhisper
- Indic Parler-TTS
- Tamil-Llama
- OpenHathi-Hi
- Sarvam Bulbul
- Sarvam Saaras
- Triton Server
- TGI
- SGLang
- TensorRT-LLM
- llama.cpp
- Ollama
- LM Studio
- Text Embeddings Inference
- Ray Serve
- BentoML
- Modal
- Cortex
- Truss
- Baseten
- LiteLLM
- OpenRouter
- DeepInfra
- Anyscale
- PyTorch
- PyTorch Lightning
- TensorFlow
- JAX
- Keras
- Transformers
- DeepSpeed
- Megatron-LM
- Accelerate
- PEFT
- TRL
- Axolotl
- Unsloth
- LLaMA-Factory
- scikit-learn
- NumPy
- Pandas
- SciPy
- Jupyter
- MLflow
- Weights & Biases
- DVC
- Optuna
- Qdrant
- Pinecone
- Weaviate
- Milvus
- Chroma
- pgvector
- Vespa
- LanceDB
- Marqo
- Faiss
- Elasticsearch
- OpenSearch
- Typesense
- Meilisearch
- Algolia
- LlamaIndex
- Haystack
- RAGAS
- PostgreSQL
- MySQL
- MariaDB
- SQLite
- MongoDB
- Redis
- ClickHouse
- DuckDB
- Snowflake
- Databricks
- Neo4j
- InfluxDB
- TimescaleDB
- CockroachDB
- Cassandra
- Apache Spark
- Apache Kafka
- Apache Airflow
- Apache Flink
- Apache Beam
- Apache Hadoop
- Trino
- Kubernetes
- Docker
- Podman
- Helm
- Istio
- Linkerd
- Argo CD
- Rancher
- K3s
- OpenShift
- Nomad
- Terraform
- OpenTofu
- Ansible
- Puppet
- Vault
- Consul
- Packer
- Pulumi
- Crossplane
- Flux CD
- ArgoCD Workflows
- NGINX
- HAProxy
- Caddy
- Envoy
- Traefik
- Kong
- Cloudflare
- Fastly
- Akamai
- WireGuard
- OpenVPN
- Tailscale
- Prometheus
- Grafana
- OpenTelemetry
- Jaeger
- Loki
- Sentry
- Datadog
- New Relic
- Honeycomb
- Langfuse
- LangSmith
- Arize Phoenix
- Helicone
- PostHog
- FastAPI
- Flask
- Django
- Express
- NestJS
- Hono
- Fastify
- tRPC
- GraphQL
- gRPC
- Apollo
- Hasura
- PostgREST
- Spring Boot
- Python
- TypeScript
- JavaScript
- Rust
- Go
- C++
- C
- Java
- Kotlin
- Swift
- Dart
- Ruby
- PHP
- Scala
- Elixir
- Erlang
- Lua
- R
- Julia
- Haskell
- Zig
- Bash
- Node.js
- Deno
- Bun
- pnpm
- npm
- Yarn
- Maven
- Gradle
- Poetry
- uv
- Next.js
- React
- Vue
- Svelte
- Solid
- Astro
- Nuxt
- Remix
- Vite
- Turbopack
✱ Open-weight models, open-source serving, sovereign hosting · no foreign API in the data path.
§ 04The India moment
The timing is not an accident.
India is no longer an AI consumer waiting for tools. The country runs one of the largest AI user bases on earth, the IndiaAI Mission has put tens of thousands of GPUs into national hands, and a generation of sovereign Indian models has arrived.
At the same time, the DPDP Act and its 2025 Rules have made data residency a board-level question, not an IT footnote.
The organisations that win the next decade will be the ones whose AI runs closest to their data — not the ones whose intelligence lives in a hyperscaler's overseas region. Marxen exists to build exactly that.
§ 05Proof
We do not pitch capability. We point at deployments.
Five government museums. An airport authority. Brand activations across hundreds of campuses. Live B2B platforms in healthcare and law. Every claim on this site is attached to something already built and running.
See our work →- Aviation● Live
Airports Authority of India · Chennai
Digital transformation work — paper-based operational logs converted to structured digital, plus passenger-facing interactive installations.
- Live B2B● Live
Naadhi · Nirnya · Vidhya · Koushalam
Four sovereign B2B platforms in production — a hospital management system, legal AI, university AI, and a skill-development platform — all on Marxen's own stack.
§ 06Start the conversation
If your data matters, where it runs matters.
Start with a conversation. We'll map your use case, your data estate, and your regulatory constraints — and tell you honestly whether sovereign AI is the right move.