Ignite AI Solutions

Track Record

Outcomes, not activity.

Ignite AI Solutions was founded in 2024. The record below is Danilo Sousa's, built leading AI and data practices before and alongside founding the company.

These engagements were delivered by Danilo Sousa as an employee or practice lead at the organizations listed, not as work performed by Ignite AI Solutions. They're shared as founder track record, not client references, because it's the standard every Ignite engagement is held to.

  1. 2025 – Present

    Avenue Code

    Senior AI Solutions Architect · AI Practice Lead

    • Built the firm's AI Solutions practice from greenfield: charter, delivery methodology, reference architectures, pre-sales playbook.
    • Drove ~$8M in AI pipeline with ~55% discovery-to-POC conversion.
    • Closed 3 production engagements totaling ~$4.8M TCV in the first 12 months.
    • Set firm-wide standards for LLM observability, evaluation and AI cost modeling.

  2. 2024 – 2025

    Marlabs

    Lead Data Scientist · AI Chapter Lead

    • Founded and led the AI Chapter from zero: 9 data scientists and ML engineers across 4 enterprise clients.
    • Reduced p95 LLM inference latency from 2.4s to 780ms and cut GPU spend by 38% (vLLM, AWQ quantization, dynamic batching).
    • Shipped 11 production models in 8 months; embedded RAG and multi-agent capability into 3 client products.
    • Established the retrieval evaluation playbook adopted as the firm's standard.

  3. 2023 – 2024

    Dana Incorporated

    Lead Data Scientist · Forecasting Tech Lead

    • Owned the LATAM S&OP forecasting roadmap across 3,200+ SKUs and 5 plants.
    • Improved forecast MAPE from 27% to 15% on critical SKUs.
    • Drove an estimated $2.3M reduction in inventory carrying cost in the first 12 months.
    • Standardized the feature engineering and backtesting framework adopted by the broader analytics team.

  4. 2016 – 2023

    John Deere

    Tech Lead, ML/DL & MLOps · Senior Data Engineer

    • Built the MLOps foundation from scratch on Vertex AI Pipelines and MLflow; cut retraining cycles from 3 weeks to 2 days.
    • Led delivery of 14+ production models spanning computer vision, anomaly detection and predictive maintenance.
    • A computer vision quality inspection model reached 92% precision and 88% recall, replacing 60% of manual inspection on 2 lines.
    • Migrated 40+ legacy ETLs to cloud-native architecture, cutting infrastructure cost ~32% and SLA breaches from 9% to under 1%.