This is what you'll be able to build in a few months.

Earnings Sentiment Analyser

by Natanel · Class of 2026

Does how a CEO sounds predict where the stock goes? Pulls earnings-call transcripts from SEC EDGAR, scores CEO confidence and guidance with Claude AI, correlates it with real post-call stock returns and surfaces everything in a live Next.js dashboard with alerts and a leaderboard.

What this proves you can doEvent-driven microservice architecture, LLM integration, financial data pipelines, cloud deployment with infrastructure-as-code, CI/CD.

DocPilot (PDF Copilot)

by Vasiliy · Class of 2026

Fill in any bureaucratic PDF form with an AI copilot. Instead of asking a chatbot and copying answers by hand, DocPilot ties each AI answer to real coordinates on the page so the answer and the place you act on it are the same interface.

What this proves you can doAI-native UX design, document processing, full-stack app deployed to production.

CodeChat

by Xiao · Class of 2026

An offline-first AI coding assistant that runs entirely on your machine with Ollama and open-source models: chat with a model, write code in a Monaco editor, switch modes, save conversations locally. No API costs, no code leaving your computer.

What this proves you can doLocal LLM integration, privacy-first architecture, rich editor UI, typed full-stack development.

ApplyLess

by Aleksei · Class of 2026

AI-powered job-matching platform for Israeli tech candidates. Automatically ingests jobs from company career pages and ATS systems, uses semantic matching to recommend the most relevant positions and generates a tailored CV for each one.

What this proves you can doData ingestion pipelines, semantic search, LLM-generated documents, product thinking.

StockTracker

by Gal · Class of 2025

A portfolio tracker for individual investors: real-time market data, personal portfolio management, performance monitoring and decision support in a clean web app.

What this proves you can doThird-party API integration, real-time data, authentication, separate front-end / back-end repos.

Budget Overflow

by Ahmad · Class of 2025

A full-stack personal finance tracker with secure cookie-based auth, category management, powerful filters and clear dashboards — coded entirely in TypeScript across the stack.

What this proves you can doEnd-to-end TypeScript, secure authentication, query builders and relational data, dashboard UI.

NeighborNet

by Elena · Class of 2025

A community action & skill-exchange hub: report neighborhood issues (potholes, noise, trash), organize to fix them together and trade skills with neighbors before involving the authorities.

What this proves you can doMulti-user social platform, CRUD + roles, map-based UI.

SeforimTracker

by Tamar · Class of 2026

A reading tracker for classical Jewish texts. Works are modeled at the canonical level with aliasing for transliteration variants and editions — a scalable foundation for bilingual search.

What this proves you can doThoughtful data modeling, stateless auth, clean separation of canonical data and user data.

SiteLens AI — Post-Event Building Damage Assessment

by Dima · Class of 2026

Reads post-disaster aerial imagery (Noto Peninsula 2024 earthquake) and returns a per-building damage assessment, rewritten for the reader: insurance adjuster, structural engineer or legal counsel. Self-assembled training set of 1,967 building crops with full provenance.

What this proves you can doBuilding a real ML dataset, image classification, retrieval-augmented reporting, rigorous evaluation methodology.

Whodunit AI — LLM-Powered Mystery Game

by Gabriel · Class of 2026

Explore locations, discover clues, interview suspects and confront them with evidence — in natural language. Behind it: a stateful LLM architecture that keeps generated dialogue grounded and consistent with what the player has actually discovered.

What this proves you can doAgentic LLM design, RAG, state management, keeping the LLM as parser while deterministic code scores.

Receipt to Cashback

by Alex · Class of 2026

Scan a paper receipt: Gemini Vision extracts items and totals in one multimodal call, FAISS matches each item to a product catalog, and a cashback engine pays users for their consumption data. A B2B analytics page adds K-Means clustering and an A/B test simulation.

What this proves you can doMultimodal LLMs, vector search, clean OOP with design patterns, statistics, containerized deployment.

Agentic Personal Finance Platform

by Victoria · Class of 2026

Upload real bank and credit-card statements (including Hebrew ones) and ask questions in plain language. An LLM orchestrates MCP tools while a custom finance MCP server does the deterministic math — reliable, explainable recommendations.

What this proves you can doAgentic architecture with tool use, MCP servers, combining deterministic logic with LLM reasoning.

InsightOS

by Raja · Class of 2026

An AI-powered desktop app that turns your document collection into a knowledge base: semantic search plus agentic file operations through natural conversation, with encrypted sandboxed file access.

What this proves you can doRAG, agentic tool execution, native desktop development, security best practices.

Fighter Predictor

by Paul · Class of 2026

Predicts UFC fight outcomes from 8,230 historical fights and 101 engineered features with an XGBoost + neural-network ensemble (73% accuracy), explains predictions with SHAP, and lets an LLM analyze similar past fights retrieved via FAISS.

What this proves you can doFeature engineering, ML ensembles and tuning, explainability, RAG-grounded LLM analysis.

Meeting Brain

by Yacine · Class of 2025

Turns meeting minutes into actions, decisions and steering KPIs, pushed directly into the team's business tools. (Transforme des comptes rendus de réunion en actions, décisions et indicateurs de pilotage intégrés aux outils métiers.)

What this proves you can doLLM extraction from unstructured text, business-tool integration, productization.

PlannerIA — Multi-agent project planning pipeline

by Michel · Class of 2025

From a one-paragraph brief to a full project plan — WBS, costs, timeline, risks and reports — orchestrated by specialized AI agents running fully on-premise on a local GPU for confidentiality.

What this proves you can doMulti-agent orchestration, local LLM deployment, API + dashboard delivery.

VERSAILLES IA

by Ludovic · Class of 2025

An immersive 3D exploration of the Petit Trianon at Versailles with an intelligent, adaptive virtual guide powered by AI.

What this proves you can doIntegrating AI, front-end 3D and back-end into one real-world application.

SaaS Subscription & Churn Analytics

by Liudmila · Class of 2025

End-to-end analytics pipeline for a subscription business: data cleaning, SQL metrics (churn, retention, MRR/ARR), hypothesis testing and a Tableau dashboard with cohort-retention heatmap.

What this proves you can doThe full analytics lifecycle: cleaning → SQL KPIs → statistics → executive dashboard.

Cardiovascular Disease Prediction

by Liza · Class of 2025

A full pipeline to analyze and predict cardiovascular disease: clinical feature engineering (BMI bins, blood-pressure categories, pulse pressure), Tableau dashboards and several ML models reaching AUC ≈ 0.80.

What this proves you can doPredictive modeling, domain-driven feature engineering, model comparison, data storytelling.

Delivery Delay Analysis & Prediction (Olist e-commerce)

by Anna · Class of 2025

Why do some Olist orders arrive late? Tableau dashboards plus a machine-learning model that predicts late deliveries and flags risky sellers and long-distance shipments.

What this proves you can doBusiness-question-driven analysis, classification models, dashboards for operations teams.

Customer Churn Analysis in E-Commerce

by Maksim · Class of 2025

End-to-end churn analysis: cleaning, feature engineering, model selection to identify the key churn drivers, and an interactive Tableau dashboard for segmentation and retention strategy.

What this proves you can doChurn modeling, customer segmentation, turning models into retention actions.

CopaAnalyzer — Copa América 2024 champion predictor

by Ariel · Class of 2024

Data analysis and an ML model to predict the 2024 Copa América champion, published as an interactive Tableau Public dashboard.

What this proves you can doSports analytics, predictive modeling, publishing a public dashboard.

Predicting Rental Apartment Prices in Israel

by Ilia · Class of 2025

Scrapes Yad2 listings, designs a relational database, trains a CatBoost price model, exposes it through a Telegram bot with a Geo API, and ships a Tableau dashboard for city-level analytics.

What this proves you can doData collection to product: scraping, modeling, ML, chatbot delivery.

AstraWatch — Monitoring Space Trends

by Rachel · Class of 2026

Transforms unstructured open-source news into a structured dataset for monitoring global space and space-security developments — ingestion via APIs, SQL analysis and Tableau dashboards by country, entity and activity type.

What this proves you can doBuilding a dataset from scratch, API integration, SQL analysis, trend dashboards.

Cyber Risk Scoring & Simulation Dashboard for SMEs

by Ilan · Class of 2025

Simulates a realistic cyber-risk scoring system for SMEs from enriched company data — composite risk score, financial-risk simulation and a professional decision-support dashboard.

What this proves you can doFeature engineering, scoring models, business-oriented dashboards.

Israel Car Market — Price Prediction Model

by Mikhail · Class of 2024

A machine-learning model that estimates the market value of used cars in Israel from historical data and specifications, with key insights on the local market for buyers, dealers and lenders.

What this proves you can doRegression modeling, in-depth EDA, market insights.

AI Compass — AI tutor agent with needs mapping & personalized ROI

by Yannick · Class of 2026

Built for a 200-person fire-safety SME whose staff already use ChatGPT without a framework: an AI agent interviews each employee, maps relevant use cases from a validated catalog, computes a personalized ROI, then activates only the training modules that matter. Business case: €22.7K pilot, +105% ROI at 12 months.

What this proves you can doNo-code agent design, tool calling, business case & ROI, responsible-AI framing.

NEXARA — Predictive SME sourcing for M&A

by Stephanie · Class of 2026

Identifies and scores French SMEs likely to be sold by analyzing weak public signals (BODACC, Pappers, Infogreffe), then produces a complete M&A memo — 28 minutes instead of 30 hours of manual work.

What this proves you can doAutomation across public data sources, scoring logic, AI-generated deliverables, honest scoping.

CoopRH Assistant — HR advice chatbot for a 750-store co-op

by Edwige · Class of 2026

Store managers get a reliable HR answer in under 2 minutes, grounded exclusively in the cooperative's official documentation with systematic source citation, sensitive-question detection and automatic escalation to HR. GDPR and EU AI Act compliant.

What this proves you can doRAG chatbot design, anti-hallucination practices, human-in-the-loop escalation.

Galence AI Watch — Automated MedTech intelligence agent

by Laurence · Class of 2026

A low-code agent continuously scans the MedTech AI ecosystem and regulatory changes, filters high-impact signals and delivers a weekly actionable digest plus ready-to-post LinkedIn content.

What this proves you can doAutomated monitoring workflows, LLM summarization, content generation with fallback plan.

Big Boss de clôture — Multi-agent monthly financial close

by Tiphanie · Class of 2026

Three specialized AI agents handle the monthly close: one writes detailed financial commentary, one builds an interactive HTML dashboard, one drafts the COMEX email from the first two — with red/yellow flags to prevent hallucinated figures.

What this proves you can doMulti-agent workflow design for finance, anti-hallucination rules, executive communication.

TAPflow AI — Transform and audit your content

by Jean-Marc · Class of 2026

An AI copilot for communication teams with three modules that automate low-value tasks, evaluate and optimize publications and support better editorial decisions — with humans kept at the center of the process.

What this proves you can doDesigning an AI copilot for a business function, human-in-the-loop workflows.

La Nițarii Inventory AI

by Eliza · Class of 2026

An internal business app that syncs catalog and stock, helps present products to customers, records sales and produces steering KPIs — connecting Base44, n8n, Google Sheets, ChatGPT, Gmail and Looker Studio.

What this proves you can doConnecting no-code tools into one operational system, turning daily operations into decision data.

Pérennis — Deal Room for business succession

by Abdul · Class of 2026

Seven AI agents orchestrated in n8n detect profitable SMEs whose owner is retiring without a successor, value them and match them with verified buyers — with a KYC/AML compliance gate and a human validating every decision. Working prototype: a Radar agent (n8n + Pappers API) and a Deal Room on Base44.

What this proves you can doMulti-agent automation, public-data APIs, compliance-by-design, MVP delivery.