Jose Antonio Jimenez

Knowledge Management & AI for BID:

Phase 1 — Strategic AI Audit and Use Case Definition

• Led AI capabilities audit as External AI Expert applying TDSP (Team Data Science Process) methodology to the BID department (commercial proposals), with executive sponsorship at director level for an aerospace and defense services organization

• Delivered Project Charter with 6+ prioritized use cases, AS-IS / TO-BE gap analysis, technical feasibility matrix, and business impact estimation, establishing the technical and economic foundation for the implementation phase

• Coordinated collaborative workshops with PM and executive stakeholders, translating business pain points (manual document search, content reuse rate <10%, multi-jurisdictional compliance complexity) into a phased TDSP roadmap

Phase 2 — Atexis Knowledge Management Platform & AI for BID Implementation

• Architected multidimensional classification system across 5 independent regulatory axes —Corporate Confidentiality, Personal Data (GDPR including Special Categories and India PDPB), National Security, Export Control, and Contractual/IP/NDA restrictions

• Implemented LLM-based classification engine using Mistral Large via Mistral Studio with per-axis decision trees, confidence scoring with human-review thresholds, and rule-based augmentation for regulatory keywords (ITAR, EAR, GDPR Article 9)

• Designed multi-collection vector storage with 5 physically segregated collections by classification level, applying chunk-level classification inheritance using 1024-dim embeddings on pgvector

• Developed CFT (Call For Tender) Analysis module: tender document parser, requirement extraction engine, semantic matching against past proposals filtered by user authorization and NDA compatibility, automated gap analysis (covered/partial/missing), pre-filled proposal template generation with inherited classification, and win probability estimation

• Architected comprehensive data science strategy for pharmaceutical manufacturing operations, designing advanced machine learning frameworks that integrate regulatory compliance (FDA, EMA, GxP) requirements with predictive analytics to drive operational excellence and innovation across production environments

• Engineered predictive modeling platform for Lot End Prediction using time-series forecasting algorithms and process analytics, delivering accurate batch completion forecasts that optimized operational planning, reduced resource waste by 15%, and improved manufacturing efficiency through real-time production insights

• Built enterprise-grade Multi-Agent LLM System using LangChain framework with locally-deployed open source models (Llama 3.1, Mistral 8x7B/8x22B), architecting specialized agents (Query Router, NER Processing,

Knowledge Graph) through LangGraph orchestration, and implementing cost-effective local inference servers with vLLM and llama.cpp for secure pharmaceutical intelligence processing • Built enterprise-grade Multi-Agent LLM System using LangChain/LangGraph framework with locally-deployed open source models (Llama 3.1 70B, Mistral 8x7B/8x22B), architecting specialized agents (Query Router, NER Processing, Knowledge Graph, Synthesis) and implementing cost-effective local inference servers with vLLM and llama.cpp, achieving 94% cost reduction vs GPT-4 while processing ~10K queries/day with <2s latency

• Designed protocol-based agent interoperability architecture: MCP-aligned tool integration patterns for secure context passing to vector databases, knowledge graphs, and external APIs; A2A-compatible coordination implementing capability discovery via query classification, structured task lifecycle management through LangGraph state machines, and standardized message schemas for inter-agent communication in regulated pharmaceutical environment

• Developed machine learning solutions for manufacturing optimization

▪ Defined the AI methodology for AI - Team Data Science Process (TDPS), aligning TDSP lifecycle phases with internal standards

for CI/CD, storage layers (Bronze/Silver/Gold), and S3 structuring.

▫ Created and maintained multiple data reports and architecture documentation, including configuration files and dataset

contracts under TDSP standards.

▫ Supported synchronization of tasks among teams, and formalization of development methodology for onboarding and

scalability.

▪ Developed reusable Kubeflow pipelines and explored MLflow deployment alternatives, integrating local MinIO simulation and

Docker Compose setups.

▪ Defined the official AI training platform integrating emitter disentanglement models, waveform generation, model training, and

inference pipelines over OpenShift AI.

▪ Led the review and QA of code repositories and pipelines, identifying maintainability risks and proposing modular orchestration

and API-based architecture.

▪ Provided continuous technical support and coordination for SuperPOD GPUs infrastructure, including resource availability,

execution tracking and SLA definition.

▪ Coordinated integration with Waveblue/Umbrella platforms, ensuring compatibility, secure data flow, and interoperability.

▪ Participated in sensitivity analysis of data flows and setup of secure development environments.

▪ Contributed to dataset standardization and mapping, helping unify the generation and transformation workflows for data.