Hooman Esteki hoomanesteki

I build AI that shows its work. AI engineer, ML engineer, software engineer, data scientist.

Data Scientist and AI/ML Engineer.
I started in analytics and product, so every system I build starts with a business question
and ends with proof: a source, a number, or a person who approves it.

_{Canadian. Open to relocate anywhere in Canada and the US.}

Right now

I build agentic AI that catches problems early.

👁️ See it	📈 Tune it	🤖 Act on it
Computer vision finds defects in images and data.	Statistics and ML set the alert line, so an alarm means something.	An AI agent weighs the risk and tells the right person what to do.

Ten problems. Ten systems that hold up.

_{Each one starts with a business problem and ends with a number. Click a name for the 15-second story.}

Project	What it does	Result
🔥 Firebreak _NEW	An AI on-call engineer. Finds what broke, or says it can't.	Every claim checked against the data
🧪 TraceGym	Catches an AI agent's silent mistakes before they ship.	10/10 planted bugs blocked
🧰 ToolSmith	Picks the right model for each job inside an agent.	3.6× cheaper per solved task
🧵 Skein	Answers with sources, or hands off to a person.	85.9% routing accuracy
🧭 Insightful Platform	Agents turn public conversations into personas you can defend.	0.98 stability score
📞 Misrouted-Call Tagger	Decides which misrouted sales calls earn a refund.	Precision 0.43 to 0.89
💬 OmniAgent	Plain-English questions in. Charts, SQL and stats out.	91% routing accuracy
♻️ Real-Time ML Platform	Notices drift, retrains and redeploys itself.	5-stage self-healing loop
🔎 Hybrid Search and RAG	Keyword and meaning search, fused into one.	+22% relevance, A/B tested
🛡️ Symbiote Lite	An AI analyst that asks before it touches your data.	0 queries run without approval

_{See all ten side by side on esteki.ca}

What I bring to a team

Craft	Tools
AI engineering	`LLM agents` `RAG` `MCP` `LangGraph` `LangChain` `LLM evaluation` `Guardrails`
ML engineering	`PyTorch` `Computer vision` `scikit-learn` `XGBoost` `MLflow` `Drift monitoring`
Data science	`Statistics` `A/B testing` `Causal inference` `Clustering` `Forecasting` `Threshold optimization`
Software engineering	`Python` `SQL` `FastAPI` `Docker` `CI/CD` `Spark` `dbt` `Databricks` `AWS` `Azure`

How I work, explained by my weekends.

Hiking _{The best talks happen uphill.}	Strength _{Showing up beats motivation.}	Running _{Long runs untangle hard problems.}
Tennis _{Lose a point. Reset. Next.}	Skiing _{Commit once the data is in.}	Polo _{Read the field. Trust the team.}