Silver Touch Technologies Secures Order to Build India's First AI Platform for Infrastructure Project Appraisal from RITES
Silver Touch Technologies Limited has secured an order from RITES Limited to develop PARAKH, India's first AI-based DPR Appraisal & Intelligence Platform, targeting the ₹111 lakh crore infrastructure sector. The self-hosted platform deploys three large language models on NVIDIA A100 GPU infrastructure with 500+ codified engineering validation rules, zero external data egress, and a Four-Layer Hallucination Prevention Framework. The order was won through the GeM portal via a competitive QCBS evaluation.

*this image is generated using AI for illustrative purposes only.
Silver Touch Technologies Limited has been awarded a significant order by RITES Limited to build India's first AI-based Detailed Project Report (DPR) Appraisal & Intelligence Platform, named PARAKH. The order covers the design, development, and implementation of the fully self-hosted, multi-model AI platform for a Government of India Navratna enterprise. This initiative targets India's ₹111 lakh crore infrastructure sector, spanning railways, highways, bridges, ports, airports, metro, and tunnels.
The project entails architecting a ground-up AI-native system rather than a conventional software application with AI features. The platform incorporates three Self-Hosted Large Language Models — Llama 3.1 70B, Mistral 7B, and Qwen 2.5 7B — deployed on NVIDIA A100 GPU infrastructure within RITES' secure data centre, ensuring zero external data egress. Additionally, the system includes 500+ Codified Engineering Validation Rules derived from IRC, RDSO, IS, and MORTH standards to deliver deterministic, auditable checks.
Technical Specifications and Security
The platform utilises a Hybrid Retrieval-Augmented Generation (RAG) combining semantic vector search with keyword search across 1,000+ historical DPRs. A Four-Layer Hallucination Prevention Framework ensures that every AI-generated insight is citation-mandatory, source-verified, confidence-scored, and human-approved. The Document Intelligence Pipeline is capable of processing DPRs ranging from 500 to 5,000 pages using OCR, table extraction, and engineering entity recognition.
| Component: | Description |
|---|---|
| AI Models | Llama 3.1 70B, Mistral 7B, Qwen 2.5 7B |
| Infrastructure | NVIDIA A100 GPU, Self-Hosted |
| Validation Rules | 500+ Codified Engineering Rules |
| Security | Keycloak IAM, AES-256 encryption, mTLS service mesh |
Integration features include GIS interoperability with the PM GatiShakti National Master Plan, BIM/IFC compatibility, and a Multi-Sector Knowledge Graph linking engineering parameters across eight infrastructure domains. The system adheres to OWASP Top 10 and LLM-specific security compliance standards.
Strategic Significance
Minesh V Doshi, Executive Director of Silver Touch Technologies Limited, stated that the partnership underscores the company's ability to engineer production-grade AI systems for mission-critical national applications. He highlighted that the platform's self-hosted nature and zero external data dependency are critical for maintaining security and sovereignty. The deployment creates a proprietary domain intelligence moat that deepens with every DPR processed by the system.
The order was secured through the Government e-Marketplace (GeM) portal following a competitive Quality and Cost-Based Selection (QCBS) evaluation. RITES Limited, a Navratna and Schedule 'A' Central Public Sector Enterprise under the Ministry of Railways, provides comprehensive consultancy services across transport infrastructure.
Historical Stock Returns for Silver Touch Technologies
| 1 Day | 5 Days | 1 Month | 6 Months | 1 Year | 5 Years |
|---|---|---|---|---|---|
| -1.20% | +1.68% | -3.43% | +26.11% | +143.89% | +2,254.94% |
Could the successful deployment of PARAKH trigger similar AI adoption mandates across other Government of India ministries?
Will Silver Touch leverage this proprietary domain intelligence to commercialize AI solutions for private infrastructure players?
How might the zero data egress architecture influence future government procurement policies regarding cloud-based AI services?





























