Dr. Allard

Director AI solution & Platform Architect

A&M Consulting • Full time

Sep 2025 - Aug 2026 • 11 mos

• Architected and operated a production platform of 19 AI agents — multi-agent orchestration (LangGraph, CrewAI), planning loops, memory and state management, tool and function calling, guardrails. • Built Model Context Protocol (MCP) servers integrating agents with document management, CRM, knowledge bases, and enterprise data platforms under least-privilege tool scoping. • Designed evaluation frameworks and rubrics measuring task success, tool-call accuracy, retrieval quality, and reasoning integrity (RAGAS, DeepEval) — detecting regressions, hallucinations, and cost overruns before production. • Built RAG pipelines end-to-end: embeddings, vector search (Pinecone, pgvector), chunking optimization, hybrid search, re-ranking, context management. • Directed a team of 10+ engineers, data scientists, and platform specialists; owned technical standards and architecture approval across the firm. • Role eliminated in July 2026 in a practice-wide restructuring under new leadership.

Principal AI Architect & Engineering Manager

IBM • Full time

Jan 2015 - Aug 2025 • 10 yrs 7 mos

• Fine-tuned, evaluated, and deployed large language models (Falcon 40B, FlanUL2, ColBERT) including training data curation, task formatting, and held-out evaluation design. • Built production NLP and document intelligence systems — text extraction, classification, semantic search, conversational AI, OCR and layout-aware processing — eliminating 65% of manual triage. • Developed statistical and deep learning models for forecasting, anomaly detection (LSTM + Mixture of Experts), and operational risk identification. • Established MLOps and LLMOps practice: CI/CD, evaluation pipelines, hyperparameter tuning, model monitoring, drift detection, automated retraining — reducing production incidents by 60%. • Architected enterprise data platforms: data lake and lakehouse, ETL/ELT, streaming ingestion, unified schema design; reduced batch processing latency by 70%. • Managed a global team of 10+ data scientists, ML engineers, and DevOps engineers across distributed and offshore delivery centers; improved retention 35%. • Delivered $4.4M in measured business impact and secured $3M+ in continued investment through executive business cases. • Established enterprise AI governance: model risk management, bias mitigation, explainability, and responsible AI practice in HIPAA and CJIS-regulated environments.