Charen Bahaeddine Hemmem
Your Average Software Guy_
About
-
Building with intent: I don't just want to make things work. I care about clean design, clear logic, and creating tools that actually feel natural to use.
-
Always iterating: To me, software is a continuous loop of building, breaking, and improving. I'm constantly learning and refining my craft, whether on the job or through side hustles.
-
Beyond the code: Code is just a tool. What really drives me is seeing how technology shapes the way we think, act, and grow.
Experience
Energisme
Python Developer
Let me do some cool stuff first and I will tell you about it later...
Awras AI
Open Source Project Lead
Facebook scraper — built a scraper for collecting data from Facebook.
YouTube scraper — built a scraper focused on harvesting audio datasets from YouTube for downstream training use.
Custom scrapers — developed several other targeted scrapers to round out the data collection pipeline.
Synthetic data generation — designed an agentic workflow that produces synthetic conversational data to augment the training set.
Annotation platform — developed the in-house annotation platform used to label and review the collected and generated data.
Model fine-tuning — fine-tuned Gemma 3 4B on the assembled datasets.
Data processing pipelines — built pipelines for cleaning, normalising, and transforming the textual data before training.
Ctopilot
Full Stack Software Engineer
Globalingo — real-time translation MVP: built the backend with FastAPI, using ElevenLabs for TTS and Whisper for STT. Evaluated and tested multiple model configurations to find the best balance of quality, latency, and pricing. Adapted the Chainlit framework (originally built for chat apps like ChatGPT) to fit the real-time translation use case. Deployed with Docker Compose and Nginx as a reverse proxy.
Regumint — AI-first legal analysis platform for EU law: built a data-intensive frontend with Next.js, React Query, and Zustand for state management. Optimised the FastAPI backend and the agent logic built with LangGraph. Deployed with Nginx as a reverse proxy.
Jawab — AI platform for learning Arabic: built the conversational agent using FastAPI and Chainlit. Forked Chainlit and edited its internals to fit the exact requirements of the product. Used PostgreSQL for persistence and MinIO for object storage. Fronted the stack with Nginx as a reverse proxy. Integrated Langfuse for prompt management and traceability.