Guumel makes AI systems production-ready — observability, evals, cost control, and the data foundation that holds underneath. Fifteen years of pipelines before the first LLM, and systems that move real metrics — not slide decks.
Databricks · Snowflake · Azure · GCP · AWS
LangGraph · LlamaIndex · LangSmith · Phoenix
Not just a vendor
Snapshot
Guumel Operating Canvas
Focus areas
01
Agent reliability: evals, observability, cost control.
02
Data foundations for AI, with Canadian data residency.
03
Data & AI strategy aligned with business outcomes.
04
Fractional data & AI leadership.
Typical engagement
2–4 wk audits · 6–12 wk builds
Outcome
Systems that keep delivering after we leave
Guumel combines deep hands-on experience in data engineering, analytics, and AI with a pragmatic consulting approach. We operate as an extension of your team — translating ambiguity into concrete architectures, roadmaps, and shipped solutions.
Whether you're hardening an AI agent that already ships, modernizing your data foundations, or clarifying your AI strategy, you get senior-level thinking from day one, not a rotating cast of juniors.
Guumel is led by Djiby Sow, AI Agent Engineer & Data Architect, based in Ottawa. Fifteen years across data platforms for financial services and insurance, and production AI agents since 2024. English and French.
Flexible, outcome-driven engagements that connect strategy, architecture, and delivery. Each project is tailored, but typically blends several of these pillars.
Agents • Data • Strategy • Advisory
01 • Agent reliability
For teams that have already shipped an AI agent or copilot and now need to prove what it does, control what it costs, and keep it from going rogue. Fixed-scope, 2–4 week engagements.
02 • Foundations
Design and implement modern data foundations so your teams can trust the numbers and your agents have something solid to reason over.
03 • Strategy
Clarify where data and AI can create real leverage and design a pragmatic roadmap that fits your business, people, and constraints.
04 • Advisory
Ongoing strategic support as a fractional head of data/AI, keeping initiatives coherent, grounded, and moving.
Names are anonymized; details available in conversation. The common thread: practical delivery and measurable impact.
Data platform
Designed and implemented a unified data layer for a multi-site clinic network, giving leadership real-time visibility into capacity, wait times, and bottlenecks.
Result: clearer prioritization, better resource allocation, and more predictable patient experience.
Agents in production
Designed and deployed a LangGraph-based agent handling text and image inputs over SMS, with state management, tool orchestration and continuous evaluation in production.
Result: 99.9% uptime, sub-second responses, and a 40% cut in token spend at equal output quality.
Analytics & automation
Helped a boutique firm standardize their data, build core dashboards, and introduce targeted automation for repetitive knowledge work.
Result: better visibility, less manual work, and more time for high-value client delivery.
RAG in a regulated domain
Built a four-agent system over a vector knowledge base to draft, validate and review regulated documents across jurisdictions, with encryption, audit logging and evaluation tooling throughout.
Result: 95% accuracy on compliance checks and a 60% shorter review cycle.
Essays and practical frameworks on building data & AI systems that create durable advantage instead of short-lived hype.
Blog coming soon – ask for drafts when we speak.
Perspective
Why so many data initiatives stall at reporting—and how to move towards genuine decision and action loops.
Coming soon
Framework
A simple way to separate shiny demos from initiatives that will meaningfully shift your P&L and customer experience.
Coming soon
Field notes
What tends to work, what tends to go sideways, and how to navigate both with minimal drama.
Coming soon
Let's work together
Pick a 30-minute slot directly, or send a short note about your context and the outcomes you care about.
Calls are in English or French. No pitch deck required — bring the problem.