Hassan M.
Khalil

AI Engineer & Data Scientist

Monty Mobile

• Built the ML core of an AI recruitment platform — automated CV parsing, embedding-based semantic scoring, and candidate ranking — cutting screening time 92%, from 12 minutes to under 1 minute, and saving 200+ hours a month.
• Designed and ran the platform's MongoDB data layer: document schemas, compound and multikey indexing, aggregation pipelines for scoring and ranking, and batched ingestion and ETL with schema validation. One documented operational store now serves both the recruiter app and the RAG assistant, with ranking-query latency cut through index and pipeline optimization.
• Surfaced shortlist-quality failure modes in the live pipeline, built a labeled ranking benchmark with precision@k and recruiter-agreement metrics, and ran champion–challenger A/B evaluation that lifted shortlist precision before rollout.
• Built an autonomous eSIM support voice agent with multi-step tool use, real-time call routing, and escalation and fallback logic (Twilio, Ultravox, OpenAI) — handling time down 60%, with 24/7 coverage.
• Architected a stacked XGBoost, CatBoost and LightGBM churn-prediction ensemble, with exploratory data analysis, SMOTE-ENN resampling and threshold tuning, then delivered at-risk dashboards and retention triggers to Sales as a revenue lever.
• Built a SIM-box fraud-detection pipeline over 20M+ call records — large-scale collection, cleaning and feature engineering feeding an autoencoder anomaly-detection model, since adopted by R&D for continued research.
• Containerized and deployed scoring and ranking services (Docker, FastAPI) on AWS through GitHub Actions CI/CD with automated tests, secret scanning and versioned model artifacts, plus continuous monitoring on p95 latency, health and prediction drift.

Stack Python, PyTorch, scikit-learn, LangChain, OpenAI, Azure, MongoDB, PostgreSQL, n8n, AWS, Docker, Power BI

Founder & Lead AI Automation Engineer

Mustashari AI consulting automation startup, part-time

• Founded and led the technical direction of an AI consulting platform delivering multi-agent workflow automation for enterprise clients across sales, finance and professional services.
• Designed a benchmark-research module running five specialized agents across frontier models (OpenAI, Claude, Gemini, Perplexity); defined the scoring rubric, comparison protocol and source-grounding checks, and analyzed cross-model disagreement to select the highest-signal outputs — turning reports that took days into ones that take under 30 minutes.
• Built an evaluation harness measuring task success rate, tool-call accuracy, groundedness and robustness across agent workflows, and wrote the internal benchmark report the team adopted as its model-selection reference.
• Shipped a suite of consulting agents (benchmark, gap analysis, deck reviewer, PowerPoint generator) and 35+ automation pipelines — RFP and RFI drafting, PM reporting, Git to LLM to docs — with REST API and webhook integrations.

Stack Python, FastAPI, LangChain, LangGraph, OpenAI Deep Research, n8n, PostgreSQL, Supabase, Docker

Contract AI & Automation Engineer

MK Energies, OTS and other clients

• Delivered end-to-end AI and automation for production use — LLM and RAG pipelines, ML models, and REST API integrations on Python and FastAPI backed by MongoDB and PostgreSQL.

Research Assistant, ML & Data Engineering

American University of Beirut

• Built reproducible ETL and feature pipelines and trained and validated ML models for wind-turbine control analysis over high-frequency multi-sensor time series, delivered to the principal investigator as a structured research deliverable; optimized PostgreSQL query performance with composite indexing.

Machine Learning Intern

Amazon ML Mentorship Program

• Selected for a competitive five-month mentorship under Amazon engineers; built and validated production-grade ML models on automated AWS data pipelines running S3 into SageMaker.