Home/References/iM Global Partner
FinTechAIAsset ManagementOn-PremiseRAG

iM Global Partner
AI RFP - Automated requests for proposals

On-premise AI platform for automatic RFP processing for a global asset manager ($38B AUM). RAG architecture, Mistral-Nemo, structured extraction and contextualized generation.

AI integration + On-Premise MVP - June 2024 - March 2025

$38B

assets under management

+50%

RFPs processed

+30%

success rate

9 months

of development

On-Premise

100% sovereign

Why this project matters

Automating RFPs in asset management

For a global asset manager handling $38 billion, each request for proposal represents hours of manual compilation. RAG AI transforms this process by generating contextualized responses in minutes, with precision and consistency impossible to achieve manually.

Sovereign AI

Mistral-Nemo 12B deployed locally with 128K token context. Complete RAG architecture: vector embeddings, semantic retrieval, few-shot generation. No financial data leaves iMGP servers.

Measurable business impact

+50% volume of RFPs processed and +30% success rate on proposals. AI standardization ensures response consistency across 16 global offices.

$38B AUM - 16 global offices - Backers: Eurazeo, Amundi, IK Partners, Luxempart
4 key modules

End-to-end AI pipeline

Document RAG

RAG (Retrieval-Augmented Generation) architecture: vectorization of historical RFP documents, semantic matching with financial products and investment strategies.

Contextualized generation

Automatic generation of RFP responses contextualized by product, geography and strategy. Advanced prompt engineering (few-shot, chain-of-thought).

Structured extraction

Automatic extraction of investment parameters into structured JSON: ticket, asset class, geography, horizon, ESG constraints, expected track record.

On-Premise deployment

100% on-premise infrastructure with Mistral-Nemo (12B, 128K context). No data leaves iMGP servers. Full GDPR and financial regulation compliance.

Measurable results

Before / After AI

MetricBefore (manual)After (AI RAG)
Volume of RFPs processedBaseline+50%
RFP success rateBaseline+30%
Compilation time per RFPSeveral daysA few hours
Cross-site consistencyVariableAI-standardized
Data securityThird-party cloudOn-Premise 100%
Deployment architecture

From audit to AI pipeline

9 months to design, deploy and optimize an on-premise AI solution that transforms the RFP process of a global asset management leader.

01

RFP process analysis

Audit of existing RFP response workflows. Identification of bottlenecks and AI automation opportunities.

02

RAG architecture

Design and deployment of the RAG architecture: vectorization of historical RFPs, embeddings, semantic retrieval pipeline.

03

On-premise MVP

Local deployment of Mistral-Nemo 12B. Integration of the complete pipeline: extraction, matching, generation, human validation.

04

Optimization & production

Prompt fine-tuning, few-shot calibration on validated responses. Production deployment with success metrics monitoring.

Differentiation vs cloud solutions - sovereign on-premise AI for finance
Technical architecture

Stack & Infrastructure

Artificial Intelligence

Mistral-Nemo 12B (128K context)
Vector embeddings
RAG (Retrieval-Augmented Generation)
Prompt engineering (few-shot, CoT)
Structured JSON extraction

Back-End

Python
NLP pipeline
Internal REST API
Semantic matching
Human-in-the-loop validation

Infrastructure

100% On-Premise
Dedicated iMGP servers
GDPR-compliant
Zero data leakage
Financial regulatory compliance

Have an AI or FinTech project?

Let's talk about your ambition

On-premise AI, document RAG, business process automation - we have the expertise to deploy artificial intelligence in the most demanding environments.