Government Tender for Development of Indigenous Gujarati Large Language Model (LLM)

This government tender invites qualified IT and AI development firms to submit proposals for the design and development of an indigenous Large Language Model (LLM) tailored specifically for the Gujarati language. Organized by Gujarat Informatics Limited, a government enterprise, this project aims to enhance language processing capabilities and promote local language technology solutions. The tender targets organizations with expertise in natural language processing, machine learning, and AI development, particularly those experienced in creating large-scale language models. The project scope includes designing, training, and deploying a robust Gujarati LLM that can support various applications such as translation, voice recognition, and content generation. The initiative aligns with the government's objective to promote digital inclusivity and linguistic diversity. The tender process is open, with detailed eligibility criteria and technical specifications provided. Bidders are encouraged to review the comprehensive tender document linked below for detailed instructions and submission guidelines.

• Design and develop a large-scale Gujarati language model capable of understanding and generating natural language content.
• Data collection and preprocessing: Gather diverse Gujarati language datasets, including textual, spoken, and multimedia sources.
• Model training: Utilize state-of-the-art NLP techniques and machine learning frameworks to train the LLM.
• Evaluation and validation: Conduct rigorous testing to ensure accuracy, contextual understanding, and language fluency.
• Deployment: Implement the LLM into accessible platforms and applications.
• Support and maintenance: Provide ongoing support, updates, and improvements post-deployment.

Specification	Details
Language Focus	Gujarati
Model Type	Transformer-based large language model
Data Sources	Text corpora, voice data, multimedia
Training Environment	High-performance computing infrastructure
Delivery Timeline	As per tender document attached

1. Submit detailed technical proposal. 2. Demonstrate relevant experience and technical capability. 3. Provide a project implementation plan. 4. Ensure compliance with data privacy and security standards. 5. Complete the project within the stipulated completion period.

✓ Must have proven experience in developing large language models or NLP solutions.
✓ Must possess relevant certifications and technical qualifications.
✓ Must demonstrate successful completion of similar projects.
✓ Bidders should have a valid Digital Signature Certificate (DSC) as per IT Act 2000.

Qualification Requirement	Details
Experience	Minimum 3 years in NLP/AI projects
Certification	Valid Digital Signature Certificate (DSC)
Financial Capacity	Annual turnover of at least ₹5 crore in the last 3 years
Technical Capability	Proven expertise in transformer models, machine learning, and data processing

**Important:** Bidders must attach scanned copies of all relevant certificates and documents as specified in the tender document.

The proposed Gujarati LLM should adhere to the following technical parameters:

Parameter	Specification
Model Architecture	Transformer-based, scalable
Data Volume	Minimum 10 million Gujarati text samples
Training Hardware	GPUs with at least 32 GB VRAM
Model Size	Not less than 1 billion parameters
Output	Fluent, contextually accurate Gujarati language generation

Sequential process: 1. Data collection and cleaning. 2. Model architecture design. 3. Model training and tuning. 4. Validation and testing. 5. Deployment and integration.

The estimated budget for this project ranges between ₹50 lakh to ₹1 crore, covering data acquisition, model development, training infrastructure, and deployment. The detailed cost breakdown is as follows:

Cost Component	Estimated Cost
Data Collection & Processing	₹10 lakh
Model Development & Training	₹30 lakh
Hardware & Infrastructure	₹15 lakh
Deployment & Support	₹5 lakh

Payment terms are as follows: - 30% advance upon contract signing. - 40% after completion of model training. - Remaining 30% post successful deployment and acceptance. All payments are subject to submission of progress reports and deliverables as per the project milestones.

1. Interested bidders must download the tender document from the official portal.
2. Submit the technical and financial bids electronically before the deadline.
3. Ensure all required documents and certificates are uploaded.
4. Bidders may attend a pre-bid meeting scheduled as per the timeline.
5. The evaluation process will commence immediately after bid submission closes.

Timeline	Date & Time
Bid Submission Start	02-06-2025 14:30
Bid Submission End	25-06-2025 15:00
Evaluation & Shortlisting	26-06-2025 to 30-06-2025
Final Selection & Award	01-07-2025

**Important:** Bidders must ensure compliance with all submission guidelines and deadlines to avoid disqualification.

Bids will be evaluated based on the following criteria:

Criteria	Weightage
Technical expertise and experience	40%
Proposed methodology and innovation	30%
Cost competitiveness	20%
Past project success and references	10%

Minimum qualifying marks for technical evaluation: 70%. The highest-scoring bid will be awarded the contract. The evaluation panel will consider the bidder’s experience, technical approach, compliance with specifications, and financial proposal. Bidders must meet all eligibility and technical criteria to qualify for the award.

Date	Event	Details
02-06-2025	Tender Publication	Tender available for download from official portal
25-06-2025	Bid Submission Deadline	Last date for bid submission at 15:00
26-06-2025	Evaluation Commencement	Evaluation process begins
01-07-2025	Award Notification	Contract awarded to successful bidder
05-07-2025	Project Commencement	Expected start date for project execution

**Note:** All dates are subject to change; bidders are advised to regularly check official updates.

For queries and clarifications, bidders can contact:

Contact Person	Designation	Contact Details
DGM (App)	Gujarat Informatics Limited	Phone: 07923255950
Email	dgmapp-gil@gujarat.gov.in

Address: GUJARAT INFORMATICS LIMITED, Block No. 2, 2nd Floor, C & D Wing, Karmayogi Bhavan Sector - 10 A, Gandhinagar, Gujarat. Bidders are encouraged to seek assistance during the vendor training sessions or via email for technical support related to bid submission.

This tender is part of the government’s initiative to promote local language technology solutions and digital inclusivity. Bidders are advised to review the detailed tender document linked below for comprehensive instructions, terms, and conditions. All submissions must adhere strictly to the specified formats and deadlines. The project aims to foster innovation in natural language processing for regional languages, with a focus on Gujarat's linguistic diversity. Participation from organizations with proven expertise in AI, NLP, and large language model development is highly encouraged. Ensure compliance with all eligibility and technical requirements to enhance your chances of selection.