AI engineer · platform builder · researcher

I help serious teams ship AI
that works in production.

I help ambitious teams turn AI into reliable products, platforms, and operating leverage. I work across architecture, LLM Ops, evaluation, safety, and delivery.

5+years building AI
94.4%published ML accuracy
15+engineers mentored
Saviour Henry and the GeoPredict AI team receiving the NCDMB award
01NCDMB
GRAND
FINALE
WINNER
2026 · ₦25M
Recent winGeoPredict AI · Team Lead
Current focusLLM Ops & AI platforms
RecognitionNational award · published researcher
Read the story

01 / 05

Not just impressive demos.
Systems people can depend on.

My work sits at the platform layer beneath AI products: model access, retrieval, evaluation, observability, safety, and developer experience. I care about the paved road - the reliable defaults that let teams move faster without losing quality or control.

02 / 05

Selected work.

01National award · 2026

GeoPredict AI

An AI-powered platform analysing seismic, well-log and core data to accelerate reservoir characterisation and drilling decisions. My team placed 1st in the NCDMB Technology Innovation Challenge, 2026.

02Stealth · Founding AI / Platform Engineer

Idiongo

A private, multi-tenant workspace where organisations talk, learn, and grow. I own the AI platform beneath the product: provider routing, workspace-isolated RAG, prompt governance, Sentinel moderation, AI course generation, Tutor, and Care Triage. Selected details are shown with permission while the product remains in stealth.

LLM GatewayRAGSentinelAI Course Generator
03Production · RAG

LegalRAG

Retrieval infrastructure rebuilt around evaluation and citations, lifting accuracy from 29% to 96.7% with structured logging, cost tracking, and full test coverage.

04Published research

CDEPF

Cross-domain ensemble probability fusion for phishing detection, reaching 94.4% accuracy and a 64.3% relative gain over baseline.

05AI systems · automation

InsightFlow

Communication intelligence that turns emails and meeting transcripts into structured, decision-ready briefs with an AI-optional, rules-first pipeline.

WORK WITH ME

Choose the kind of
leverage you need.

03 / 05

Experience that
compounds.

2025 - now

Turing AI/ML Engineer · LLM Evaluation & Alignment

Built automated multi-dimensional evaluation frameworks, agentic code-repair pipelines reaching 85% SWE-Bench success across 200+ issues, and SFT/RLHF data pipelines with researchers.

2025 - now

Stealth B2B SaaS Founding AI / Platform Engineer

Own the shared AI platform layer: provider failover, RAG infrastructure reused across five AI surfaces, prompt versioning, generation tracing, observability, moderation, and a support agent resolving ~42% of tickets at ≥0.92 confidence.

2024 - 2026

Freelance Senior AI Engineer · Evaluation & Platform Tooling

Shipped SwarmBench evaluation environments, regression pipelines in GitHub Actions, agentic workflows, and research-intelligence RAG systems with citation tracking.

2024 - now

Start Innovation Hub Nigeria Machine Learning Engineer

Architected NLP/CV systems processing 50K+ documents and images monthly, cut inference latency 70%, and mentored 15+ engineers. Staff of the Year ×2.

2022 - 2024

Visis Startup · Presprint Digital Machine Learning Engineer

Production NLP/CV, OCR, anomaly detection, deployment workflows, monitoring, and MLOps foundations across accessibility and predictive systems.

04 / 05

Research,
recognition, range.

Published · Franklin Open (Elsevier) · 2026

Ensemble Transfer Learning for Cross-Dataset Phishing Detection

CDEPF framework: 94.4% cross-dataset accuracy through PCA harmonisation and information-theoretic fusion.

01NCDMB Technology Innovation Challenge
Grand Finale Winner · GeoPredict AI
02Young Africa Innovates
Top 30 of 9,000+ ↗
03African Impact Challenge
Cohort 6 · Builder Track
04DeepTech Ready Spotlight
Mentor · DSN / Google.org / 3MTT

05 / 05

How I help teams
ship better.

01

LLM Ops & platform engineering

Gateways, provider failover, prompt registries, tracing, cost and safety controls.

02

RAG & evaluation

Grounded retrieval, reranking, citation verification, human-in-the-loop measurement.

03

ML Ops & infrastructure

FastAPI, NestJS, Docker, CI/CD, monitoring, drift detection, canary deploys.

04

Technical leadership

Architecture, research translation, documentation, and mentorship that compounds.

Let’s talk

Have a hard problem?

Let’s make it
matter.

Start a conversation

Tell me what you are building, where it is stuck, and what success looks like. The usual first step is a focused architecture conversation, then a scoped sprint or platform engagement.

johnsavilesh@gmail.comLinkedIn ↗GitHub ↗