Skip to content
View berkde's full-sized avatar
  • Boston University
  • Boston, MA, United States
  • 06:43 (UTC -05:00)

Highlights

  • Pro

Block or report berkde

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 supported. This note will be visible to only you.
Report abuse

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

Report abuse
berkde/README.md

Berk Delibalta

AI & Software Engineer | Backend • Cloud • LLMs • Reinforcement Learning

I build production-ready systems at the intersection of Backend Engineering, Cloud Architecture, and Artificial Intelligence.
My work spans:

  • High-performance Java & Python backend systems
  • Cloud-native architectures on AWS
  • Agentic AI systems (LLM tool-calling, reasoning pipelines, workflow orchestration)
  • Reinforcement Learning algorithms with C++/LibTorch
  • Scalable, intelligent applications deployed end-to-end

I’m currently completing my M.S. in Computer Science at Boston University, and previously earned a B.Eng. in Computer Engineering from Politecnico di Torino.


Software • Cloud • AI Engineering

  

LinkedIn


Featured Projects

FindYourDreamHouseAI — AI-Powered Real Estate Search

Spring Boot • Redis • Postgres • Ollama • LLM Agents • RL Ranking
A full AI-driven platform that processes natural-language queries, images, and metadata to recommend real estate listings using:

  • Multi-agent LLM workflows
  • Custom RL-based ranking engine
  • Vector search + deep embeddings
  • Secure, scalable Spring Boot backend
    ➡️ Agentic AI + backend engineering + cloud deployment

🔗 https://github.com/berkde/findYourDreamHouseAI


TestMIND — AI Test Case Generator

Python • FastAPI • React • OpenAI • CI/CD
Generates complete test suites (positive/negative/edge cases) from English requirement descriptions. Includes:

  • Intelligent parsing + reasoning pipelines
  • AllPairs combinatorial reduction
  • Speech-to-text input
  • React UI + full backend workflow
    ➡️ LLM workflow engineering + full-stack + testing automation

🔗 https://github.com/berkde/testmind


FastTD3 — High-Performance RL Algorithm

C++ • LibTorch • MuJoCo • HPC
A production-grade TD3 implementation with:

  • Eigen-based + PyTorch-based versions
  • Continuous-action humanoid control
  • Simulation acceleration with MuJoCo
  • Full CI/CD and unit testing
    ➡️ Reinforcement learning + C++ systems + performance engineering

🔗 https://github.com/berkde/FastTD3


Additional Projects

  • NYC House Price Prediction — XGBoost/RF stacked model + Streamlit app
  • Audio Classifier AI — CNN on mel spectrograms
  • TensorFlow Integer Quantization — TFLite model optimization
  • NYC Airbnb Popularity Prediction — ML ranking pipeline

Technical Areas

Backend Engineering: Java, Spring Boot, Python, FastAPI, Redis, Postgres, Kafka
Cloud Engineering: AWS Lambda, Step Functions, DynamoDB, S3, RDS, EC2, VPC
AI Engineering: LLM application design, agentic AI workflows, prompt engineering
ML/RL: TD3, PPO, continuous control, LibTorch, MuJoCo simulations
DevOps & CI/CD: Docker, Kubernetes, GitHub Actions, distributed tracing


Professional Experience

  • Deloitte — Cloud Software Engineer
  • Zurich Insurance — Software Engineer
  • Thales — Backend & C++ Engineer (Defense/Telecom)
  • IBM — Software Engineer (Watson Cloud)
  • Boston University — Reinforcement Learning Researcher

Contact


Building intelligent, scalable, and production-ready AI systems.

Pinned Loading

  1. findYourDreamHouseAI findYourDreamHouseAI Public

    A modern, secure, and scalable application for real estate management.

    Java 1

  2. testmind testmind Public

    AI-assisted test case generator using NLP, FastAPI and OpenAI

    Python

  3. FastTD3 FastTD3 Public

    This is a C++ implementation of the FastTD3 algorithm, a high-performance variant of the Twin Delayed Deep Deterministic Policy Gradient (TD3) algorithm optimized for complex humanoid control tasks.

    C++ 1

  4. NYC-House-Price-Prediction-App NYC-House-Price-Prediction-App Public

    This project predicts real estate prices for NYC properties using a stacked machine learning model that combines XGBoost and Random Forest with a Ridge Regression meta-learner. The app is served vi…

    Jupyter Notebook

  5. Deep-Learning-Post-Training-integer-Quantization-with-8-Bit-Precision-Using-Min-Max Deep-Learning-Post-Training-integer-Quantization-with-8-Bit-Precision-Using-Min-Max Public

    his project demonstrates how to train a TensorFlow model to recognize handwritten digits from the MNIST dataset and then optimize it for deployment using TensorFlow Lite. The optimization technique…

    Jupyter Notebook

  6. DeepRL DeepRL Public

    This project implements and compares major reinforcement learning TD control algorithms on the Windy Cliffwalking environment

    Python