Yasmine Elnasharty
AI Engineer
JavaScript
TypeScript
Node JS
SQL
Python
Data Pipeline & Feature Engineering: 4/10
Active 7 days ago
+20 (10) 10492714 Invite to interview
Message
Download CVCV
Overview
Technical skills
Timeline
Roles
Overview
Bioinformatics-focused ML practitioner (middle level) who builds end-to-end exploratory analyses and visualizations for protein sequences and public-health datasets. The strongest proven skill is practical domain scripting and visualization, demonstrated by the protein sequence analysis notebook functions such as fetch_protein_sequence, basic_sequence_analysis, plot_aa_distribution and the alignment/heatmap visualizations. What is not evidenced is production-grade engineering, automated testing or CI/CD, robust model validation (cross-validation/calibration), or large-scale data pipeline and distributed systems experience.
Technical skills
JavaScript
TypeScript
Node JS
SQL
Python• Middle
Node JS
Nest.JS
Python
Flask• 3y+
FastAPI
Databases
Oracle
SQLite
PostgreSQL
AI/ML
AI Agents
AWS Bedrock
Claude
Claude Code
Cursor
DeepEval
Gemini
Groq
Langfuse
LLM
Multimodal AI
Perplexity
Prompt Engineering
Ragas
Vertex AI
Jupyter Notebook
Pandas
NumPy
SciPy
Scikit-learn
LangChain• 3y+
Llama• 3y+
CrewAI
Gemma
LangGraph
LangSmith
RAG
DevOps
AWS
CI/CD
Docker
GCP
Git
Vercel
Rest API
WebSockets
Azure
Vector
Analytics
Matplotlib
Seaborn
Frontend
D3.js
Cybersecurity
GDPR
Timeline
Senior AI Engineer
•
Senior
Bold AI
•
Full-Time
Architected a multi-LLM brand-visibility engine that parallelizes probe queries across multiple model providers and unifies results with an LLM judge layer for sentiment scoring. Built a supervisor-orchestrated multi-agent research system that decomposes requests into specialist agents with isolated context and token-budget controls. Implemented deterministic and auditable agent routing with reflection-based re-dispatch, Postgres checkpointing, and human-in-the-loop approvals for high-stakes actions. Added end-to-end observability with LangSmith tracing, offline evaluation, and GDPR-aligned PII scrubbing for customer deployments.
LangGraph
LangSmith
PostgreSQL
GDPR
Senior AI Engineer
•
Senior
RecruitPilot AI
•
Full-Time
Engineered a LangGraph multi-agent system with supervisor routing, agent-to-agent handoffs, and human-in-the-loop checkpoints, exposing tools that automate candidate and job creation workflows across CRM platforms. Built autonomous agents for candidate sourcing, job search, and lead generation, improving match accuracy through iterative agent workflows. Developed a user-scoped RAG pipeline on Azure AI with document indexing and per-user filtering. Automated meeting note-taking across platforms by integrating an agentic meeting-bot orchestration directly into the LangGraph workflow.
LangGraph
Azure
RAG
AI Engineer
•
Middle
RealAIzation
•
Full-Time
Built an enterprise BI agent on LangGraph with specialized SQL, planning, summary, and chart agents, using SQLite checkpointing over an Oracle vector database. Created a multi-agent learning-path generator using dedicated query generation, retrieval, sequence planning, and validation agents. Implemented hybrid retrieval combining vector search and BM25 over a large course catalog for low-latency responses. Delivered a budget-analysis system using CrewAI plus LangGraph, streaming real-time insights via a FastAPI WebSocket interface for D3.js dashboards.
LangGraphsince 2024
SQL
SQLite
Oracle
CrewAI
FastAPI
WebSockets
D3.js
Tanta University
Bachelor's Degree •
Computer Science
AI Engineer
•
Middle
AirCampus AI
•
Full-Time
Architected an AI tutoring platform with a Flask plus LangChain microservices backend, using Groq-hosted Llama/Gemma models. Implemented real-time voice interaction by integrating an STT-to-LLM-to-TTS processing pipeline within the service architecture. Built a bilingual (Egyptian Arabic and English) voice sales agent with mid-call voice handoff behavior and deterministic persona transitions for voice roleplay training. Added an analytics layer to track learning progression and provide instructors with visibility into student engagement and mastery over time.
Flask
LangChain
Llama
Gemma
Middle AI/ML Engineer
Confidence: Medium Data-centric
Bioinformatics and data-focused practitioner at a Middle level specializing in protein sequence analysis and public-health data exploration. The strongest proven skill is practical domain scripting and exploratory analysis using Biopython-based sequence workflows and BLAST/motif/structure parsing as seen in the protein analysis notebooks and 7qih_PDB.py. There is little evidence of production ML engineering, experiment tracking, automated testing, or model serving infrastructure in public code.
Model Architecture & Training
2/10
How well models are designed and trained
Minimal model work: uses off-the-shelf scikit-learn models (logistic and linear regression) and reporting but no custom architectures, training loops, hyperparameter tuning, or validation pipelines.
Evidence
Assessing_Anemia_in_Pregnant_and_Non-Pregnant_Women/Assessing_Anemia_in_Pregnant_and_Non-Pregnant_Women.ipynb: training LogisticRegression() and using classification_report
Assessing_Anemia_in_Pregnant_and_Non-Pregnant_Women/Assessing_Anemia_in_Pregnant_and_Non-Pregnant_Women.ipynb: LinearRegression model fit and metrics table
Data Pipeline & Feature Engineering
4/10
How data is prepared for models
Clear data cleaning and feature extraction for domain data; multiple functions extract numeric hemoglobin values, dropna/duplicate handling, pivoting, and sequence file I/O for bioinformatics workflows.
Evidence
Assessing_Anemia_in_Pregnant_and_Non-Pregnant_Women/Assessing_Anemia_in_Pregnant_and_Non-Pregnant_Women.ipynb: process_hemoglobin_data, extract_numeric / extract_hemoglobin and pivot/concatenation pipelines
Exploring_Protein_Folding_with_AlphaFold/Exploring_Protein_Folding_with_AlphaFold.ipynb: load_sequence, calculate_composition, save FASTA files and read_fasta utilities
Experimentation & Evaluation
3/10
How results are measured and tested
Basic evaluation and exploratory analysis are present (confusion matrices, classification reports, Pearson correlation, p-values, visual EDA), but there is no experiment tracking, CV, ablation, or reproducibility framework.
Evidence
Assessing_Anemia_in_Pregnant_and_Non-Pregnant_Women/Assessing_Anemia_in_Pregnant_and_Non-Pregnant_Women.ipynb: confusion_matrix, classification_report and ConfusionMatrixDisplay usage
Protein_Sequence_Relationship_Analysis/Protein_Sequence_Relationship_Analysis.ipynb: calculate_pearson_correlation returning correlation and p-value
MLOps & Deployment
1/10
How models are shipped to production
Almost no MLOps or deployment evidence; only simple file saves and plotting. No model versioning, serving, CI/CD, or monitoring.
Evidence
Protein_Sequence_Relationship_Analysis/Protein_Sequence_Relationship_Analysis.ipynb: plot and save figure 'sequence_alignment.png'
Exploring_Protein_Folding_with_AlphaFold/Exploring_Protein_Folding_with_AlphaFold.ipynb: writes FASTA files to local paths
Computational Efficiency
1/10
How efficiently computing resources are used
No explicit computational-efficiency engineering such as GPU/TPU use, batching, memory profiling, quantization or distributed training; only routine use of established libraries for alignments and BLAST.
Evidence
Protein_Sequence_Relationship_Analysis/Protein_Sequence_Relationship_Analysis.ipynb: use of NCBIWWW.qblast and pairwise2/pairwise alignments (computationally heavy calls but no efficiency work)
Exploring_Protein_Folding_with_AlphaFold/7qih_PDB.py: PDBParser usage for structure parsing (no optimization)
Research Depth & Innovation
2/10
Depth of research and new ideas
Domain-aware bioinformatics scripting and use of Biopython, motif analysis and AlphaFold outputs show domain knowledge, but there is no novel research contribution, custom layers, rigorous benchmarking, or reproduced paper results.
Evidence
Protein_Sequence_Relationship_Analysis/Protein_Sequence_Relationship_Analysis.ipynb: motifs.create and motif.consensus usage
Exploring_Protein_Folding_with_AlphaFold/7qih_PDB.py: load_structure and extract_secondary_structure using Bio.PDB
Expertise
Medical AI & Healthcare• Middle
Scientific AI (Life Sciences, Physics, Climate)• Middle
Industries
Health Care• Middle
Science & Engineering• Middle
Technologies
SQL
PostgreSQL
Cursor
LangGraph
Rest API
LangChain• 3y+
Claude
GCP
Oracle
Flask• 3y+
Claude Code
Groq
Vercel
Vertex AI
WebSockets
Gemma
Prompt Engineering
Multimodal AI
AI Agents
DeepEval
Langfuse
LangSmith
Ragas
Azure
Llama• 3y+
AWS Bedrock
CI/CD
Git
SQLite
AWS
Docker
CrewAI
Gemini
LLM
RAG
Perplexity
Vector
Recommendations
- Turn notebooks into small, reusable Python modules or scripts with unit tests and argument parsing to make analyses reproducible and easier to integrate into pipelines.
- Add basic experiment tracking and reproducibility (e.g., a simple MLflow/W&B integration or structured logging) for the modeling steps to capture datasets, parameters and metrics.
- Harden data extraction code with robust parsing and error handling for real-world CSV and FASTA inputs and avoid hard-coded /content paths to improve portability.
- Package frequent bioinformatics operations (sequence fetch, alignment, motif extraction) into functions with clear inputs/outputs so they can be used in automated pipelines or notebooks.
Repositories
The developer's experience in this domain has been verified based on AI analysis of the following repositories:
Middle Data Scientist
Confidence: Medium ML Practitioner
Bioinformatics-focused ML practitioner (middle level) who builds end-to-end exploratory analyses and visualizations for protein sequences and public-health datasets. The strongest proven skill is practical domain scripting and visualization, demonstrated by the protein sequence analysis notebook functions such as fetch_protein_sequence, basic_sequence_analysis, plot_aa_distribution and the alignment/heatmap visualizations. What is not evidenced is production-grade engineering, automated testing or CI/CD, robust model validation (cross-validation/calibration), or large-scale data pipeline and distributed systems experience.
Statistical Rigor
3/10
Correct use of statistics
Basic statistical checks are present (Pearson correlation and p-value) but there is little treatment of assumptions, uncertainty quantification, multiple-comparison controls, or honest negative-result discussion; there is also at least one clear misuse where true labels are reused as predictions.
Evidence
Protein_Sequence_Relationship_Analysis/Protein_Sequence_Relationship_Analysis.ipynb: calculate_pearson_correlation
Assessing_Anemia_in_Pregnant_and_Non-Pregnant_Women/Assessing_Anemia_in_Pregnant_and_Non-Pregnant_Women.ipynb: classification_report / confusion_matrix (true labels used as predictions)
Data Wrangling & Cleaning
4/10
Preparing and cleaning data
Reasonable ad-hoc data wrangling and domain-specific parsing are implemented (FASTA handling, sequence fetching, tooltip numeric extraction) but cleaning is simple, paths are hardcoded, and there is little defensive error handling or provenance/versioning.
Evidence
Assessing_Anemia_in_Pregnant_and_Non-Pregnant_Women/Assessing_Anemia_in_Pregnant_and_Non-Pregnant_Women.ipynb: process_hemoglobin_data (extract and convert 'First Tooltip'), dropna/drop_duplicates
Exploring_Protein_Folding_with_AlphaFold/Exploring_Protein_Folding_with_AlphaFold.ipynb: load_sequence, save FASTA files
Protein_Sequence_Relationship_Analysis/Protein_Sequence_Relationship_Analysis.ipynb: fetch_protein_sequence (ExPASy/SwissProt fetching)
Exploratory Analysis & Visualization
5/10
Exploring and visualizing data
Extensive, polished visualizations and domain-specific plots (amino acid distributions, alignment visualizations, heatmaps) with annotated thresholds and styling; however narrative interpretation after figures is limited and some plots use fixed/hardcoded aesthetics.
Evidence
Protein_Sequence_Relationship_Analysis/Protein_Sequence_Relationship_Analysis.ipynb: plot_aa_distribution, plot_similarity_matrix, plot_sequence_alignment
Exploring_Protein_Folding_with_AlphaFold/Exploring_Protein_Folding_with_AlphaFold.ipynb: plot_composition, plot_hydropathy
Predictive Modeling
3/10
Building models that predict
Baseline predictive modeling is present (logistic and linear regression) and evaluation artifacts exist, but modeling is simplistic (very few features), lacks proper cross-validation, calibration, careful error analysis, and in places evaluation code is logically flawed.
Evidence
Assessing_Anemia_in_Pregnant_and_Non-Pregnant_Women/Assessing_Anemia_in_Pregnant_and_Non-Pregnant_Women.ipynb: train_test_split + LogisticRegression + classification_report
Assessing_Anemia_in_Pregnant_and_Non-Pregnant_Women/Assessing_Anemia_in_Pregnant_and_Non-Pregnant_Women.ipynb: LinearRegression sections and model metrics table
Business Insight & Impact
2/10
Turning analysis into business value
Some public-health framing is visible (anemia thresholds, regional trend plotting) but there is little translation to actionable business/policy recommendations, cost-of-error analysis, or prioritized operational impact.
Evidence
Assessing_Anemia_in_Pregnant_and_Non-Pregnant_Women/Assessing_Anemia_in_Pregnant_and_Non-Pregnant_Women.ipynb: anemia threshold vertical lines and regional regression plots
Exploring_Protein_Folding_with_AlphaFold/README.md (project objective statements used for domain framing - NON-SCORING CONTEXT used to infer domain)
Reproducibility & Notebook Hygiene
2/10
Clean, repeatable analysis
Notebooks are executable examples but reproducibility is limited: they include inline '!pip install' commands, hardcoded local paths, no pinned environment file, no seeds, and no pipeline orchestration or data versioning.
Evidence
Protein_Sequence_Relationship_Analysis/Protein_Sequence_Relationship_Analysis.ipynb: '!pip install' lines and direct system calls
Exploring_Protein_Folding_with_AlphaFold/Exploring_Protein_Folding_with_AlphaFold.ipynb: hardcoded '/content/*.fasta' and '/content/*.pdb' file paths
Industries
Biotechnology• Middle
Health Care• Middle
Science & Engineering• Middle
Technologies
Jupyter Notebook
Scikit-learn
Seaborn
Matplotlib
SciPy
Pandas
NumPy
Recommendations
- Develop exploratory bioinformatics analyses and reproducible notebooks that visualize sequence properties and alignments for small-to-medium scale studies.
- Prototype and validate simple predictive models or statistical analyses for biological or public-health datasets where careful feature engineering and domain thresholds are required.
- Create teaching or demonstration material (workshops, tutorials) that show sequence fetching, FASTA handling, alignment and visualization for bioinformatics students or researchers.
Repositories
The developer's experience in this domain has been verified based on AI analysis of the following repositories:
Intern Backend Developer
Confidence: Low API Engineer
Backend-focused developer (entry level) with exposure to building AI-driven chat and voice integrations and a strength in wiring LLM/STT/TTS services. The clearest proven skill is integrating LLM and speech services into API endpoints, as evidenced by the Python backend code that integrates LLM clients and speech transcription/TTS flows. There is little or no human-authored evidence of production-grade data modeling, migrations, resilience patterns, or structured observability in the public human-authored files.
API Design
How well APIs are designed
Not evidenced in public code
Data Layer & Database
Working with databases
Not evidenced in public code
Scalability & Performance
Handling load and speed
Not evidenced in public code
System Architecture
Overall system structure
Not evidenced in public code
Security & Auth
Protecting data and access
Not evidenced in public code
Reliability & Observability
Stability and monitoring
Not evidenced in public code
Expertise
Backend AI & LLM• Intern
Python• Intern
Node.js• Intern
Industries
Artificial Intelligence• Intern
Education• Intern
Technologies
Python• Middle
Node JS
FastAPI
Flask• mentioned only
Recommendations
- Develop REST/streaming APIs that include explicit timeout, retry-with-backoff, jitter and idempotency keys for LLM calls and external STT/TTS services.
- Implement a simple migration history and database schema (with example migration files) and show transaction boundaries and basic indexing decisions for a real feature.
- Add structured logging, request correlation ids, and a minimal metrics dashboard (Prometheus/Grafana) to demonstrate observability and incident diagnosis.
- Harden API surface: replace permissive CORS and add input validation, auth token lifecycle (refresh/revocation), and secrets handling examples (env + vault integration).
Repositories
The developer's experience in this domain has been verified based on AI analysis of the following repositories: