Skip to content
View hoomanesteki's full-sized avatar

Highlights

  • Pro

Block or report hoomanesteki

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
hoomanesteki/README.md

Typing SVG


Website  LinkedIn  YouTube  Instagram  Location


I build agentic AI, RAG, ML platforms, and data products that move from prototype to production.

My focus: grounded outputs, measurable decisions, code-level safety, observability, and human approval for high-risk actions.

🎓 Master of Data Science, UBC
☁️ AWS and Databricks Certified
📍 Open across Canada and the US

     

I work where “the model works” becomes “the system is ready to ship.”


🚀 Things I've Shipped

Seven end-to-end systems built around real problems, measurable outcomes, and practical trade-offs.

P1

🧵 Skein Lite: Agentic RAG Knowledge Platform

Agentic RAG · Knowledge Graph · Semantic Layer · MLOps

A domain-adaptable assistant that answers with evidence or abstains.

A master orchestrator routes requests to five specialists using vector retrieval, reranking, structured metrics, and a knowledge graph. Deterministic gates block PII, prompt injection, unsafe actions, and unsupported questions before the model runs.

Retraining is automated, while model promotion stays human controlled.

500 tests 85.9% routing 100% escalation recall 430× lower cost

FastAPI LangGraph Qdrant Cohere Groq Neo4j dbt MLflow Langfuse

Read the story → · Live showcase →

P2

🧬 From Posts to Personas

NLP · Clustering · Statistical Validation

Turns 83K social posts into evidence-backed personas.

Behavioral signals are clustered, then validated using Fisher or chi-square tests with false-discovery-rate correction. Each persona is supported by statistically significant patterns, not only visual clusters.

83K posts Validated personas 23K+ tests

DSPy GEPA HDBSCAN K-Means Fisher BH-FDR

Read the story →

P3

📞 Misrouted-Call Tagger

LLM Extraction · Classification · Cost-Aware Design

Classifies whether a misrouted lead should be refunded.

Moving a key decision from the model into a deterministic rule improved precision from 0.43 to 0.89, without using a larger model.

.43 → .89 precision Same model Lower cost

LLM Extraction Embeddings XGBoost Rules Shadow Mode

Read the story →

P4

♻️ Real-Time ML Platform

MLOps · Drift Detection · Continuous Training

A five-stage lifecycle connecting production monitoring to retraining.

PSI drift triggers retraining, challenger evaluation, artifact tracking, and human-reviewed promotion.

5-stage loop PSI monitoring Zero manual scheduling

FastAPI MLflow Docker AWS CI/CD Model Registry

Read the story →

P5

🔎 Hybrid Retrieval and RAG

Search · Reranking · Safe RAG

Combines BM25 and semantic retrieval using reciprocal rank fusion and reranking.

The guarded RAG layer answers from retrieved evidence, blocks prompt injection, and limits SQL to SELECT-only queries.

+22% relevance Sub-second A/B tested

FAISS BM25 SBERT RRF Reranking LangChain

Read the story →

P6

🧠 OmniAgent Analytics

Multi-Agent System · Analytics · Secure Execution

Turns natural language into SQL, charts, statistics, and predictions.

A central orchestrator routes work to seven specialists with schema validation, sandboxed execution, and 40+ security checks.

91% routing 7 agents 195+ tests

FastAPI MCP Plotly Docker SQL Python

Read the story →

P7

🛡️ Symbiote Lite

Human-in-the-Loop · SQL Safety · Fallbacks

An AI analyst that asks before it acts.

A three-stage approval flow protects sensitive operations. SQL is checked at the AST level, writes are blocked, and deterministic fallbacks keep core analytics available.

154+ tests 1M+ records Zero writes

FastAPI Gradio MCP SQLite AST Validation

Read the story →


🧭 How I Build

🛡️ Guarded

Safety rules and approval gates live in code, not prompts.

📏 Measured

Models are selected by quality, latency, cost, and failure analysis.

💸 Cost-Aware

Rules and smaller models handle simple requests first.

🤝 Human Approved

Retraining is automated. Production promotion is controlled.


🛠 Toolbelt

🧵 RAG & Retrieval

LangChain LangGraph Qdrant FAISS BM25 SBERT Cohere Neo4j RRF Reranking

🤖 Agents & Safety

MCP DSPy GEPA Groq Multi-Agent Routing Human-in-the-Loop Guardrails

🚀 MLOps & LLMOps

MLflow Langfuse RAGAS FastAPI Docker CI/CD Model Registry Drift Monitoring

🧠 ML & NLP

PyTorch scikit-learn XGBoost Transformers Embeddings Clustering

🗄️ Data & Cloud

Spark PySpark Databricks dbt DuckDB PostgreSQL AWS Azure

📊 Statistics

Python R A/B Testing Causal Inference Fisher Chi-square BH-FDR


🌄 Life Beyond the Laptop

Outside work: mountains, movement, and competition.


⛷️ Skiing

🎾 Tennis

🏃 Running

🐎 Polo

🥾 Hiking

🏋️ Strength

📊 GitHub Activity


def build_production_ml(problem):
    baseline = measure(problem)

    system = build(
        grounded=True,
        guarded=True,
        observable=True,
        cost_aware=True,
        human_approved=True,
    )

    return system.ship()

Portfolio   LinkedIn


Machine Learning Engineer · Data Scientist · Master of Data Science, UBC

Popular repositories Loading

  1. iris-ml-predictor iris-ml-predictor Public

    A machine learning workflow for classifying iris flowers using the famous Iris dataset. It provides options for both Docker and Conda-based execution environments.

    HTML 1

  2. symbiote-lite-analyst-agent symbiote-lite-analyst-agent Public

    A lightweight human-in-the-loop AI agent that augments analyst workflows through guided reasoning, synthesis, and decision support.

    Jupyter Notebook 1

  3. omniagent-multi-agent-analytics omniagent-multi-agent-analytics Public

    OmniAgent is an AI-powered conversational data assistant with a multi-agent architecture. It automatically routes your questions to the right agent to generate insights, visualizations, aggregation…

    Python 1

  4. pipecat-voice-agent-rag-mcp pipecat-voice-agent-rag-mcp Public

    Real-time voice AI agent built with Pipecat, featuring RAG knowledge retrieval and MCP tool integration. Sub-400ms latency with Deepgram STT, Claude LLM, and Cartesia TTS.

    1

  5. agentic-rag-knowledge-ai-platform agentic-rag-knowledge-ai-platform Public

    Skein is an agentic AI platform that answers questions by weaving together structured and unstructured data, grounding every response in a knowledge graph, hybrid vector search, and a governed sema…

    Python 1

  6. hoomanesteki hoomanesteki Public