Skip to content
View vladihnat's full-sized avatar
0 followers · 1 following

Block or report vladihnat

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.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

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

Report abuse
vladihnat/README.md

Hi, I'm Vladimir Herrera Nativi 👋

M2 student in Artificial Intelligence at Université Paris-Saclay, working on reinforcement learning, generative models and representation learning. I'm looking for an M2 research internship leading to a CIFRE PhD, with a particular interest in self-supervised world models (JEPA).

  • 🔬 Research interests: deep RL for control & energy systems · generative models (VAEs, recurrent/attention models) · self-supervised learning (JEPA) · computer vision
  • 🧪 Research experience: RL for microgrid energy management (Université de Corse, SPE lab — Paglia Orba solar platform) · supervised vs RL for portfolio optimization (LISN, Paris-Saclay)
  • 🎓 Background: Double licence Mathematics & Computer Science + Magistère in Computer Science (Paris-Saclay) · AEFE excellence scholar
  • 🌍 Languages: Spanish (native) · French (C2) · English (C1)

⭐ Featured projects

Project What it is Stack
rl-ems RL energy management system for a PV + battery microgrid. A SAC agent replaces a MILP optimiser under real operating constraints (no grid charging, peak/off-peak tariffs); experiments focus on closing the gap to the MILP optimum on held-out data. Stable-Baselines3 · Gymnasium · CVXPY/HiGHS
DRAW recurrent VAE Re-implementation of DRAW (Gregor et al., 2015): iterative image generation on MNIST, ablating RNN vs LSTM and with/without attention. Python · deep generative models
Climbing Coach AI Computer-vision pipeline that analyses climbing technique from video: pose estimation → biomechanical metrics (hip-to-wall distance, torso lean, arm extension) → visual feedback. OpenCV · MediaPipe · scikit-learn
Financial news NLP Frugal transformer pipeline for summarisation, entity extraction/linking and market-impact prediction on financial news (team project). Hugging Face Transformers
Pollinator detection challenge Designed and solved an ML challenge for pollinator detection (data, baseline, scoring), hosted on Codabench. Python · Codabench
Bias detection in chest X-rays Detecting and analysing bias in the NIH ChestX-ray14 medical imaging dataset. Python · Jupyter

🛠️ Tools

Python · PyTorch · Stable-Baselines3 · Gymnasium · Hugging Face · scikit-learn · NumPy/pandas · CVXPY · Git · LaTeX

📫 Contact

LinkedIn · 📍 Paris area, France

Pinned Loading

  1. rl-ems rl-ems Public

    RL-based Energy Management System for microgrids (PV + battery). Replaces a MILP baseline with a RL agent trained via Stable-Baselines3, using python-microgrid for simulation and CVXPY/HiGHS for th…

    Python 1 1

  2. financial-news-nlp financial-news-nlp Public

    This repo is a self-contained project exploring diverse NLP tasks on a High-Quality Financial News Dataset.

    Jupyter Notebook

  3. draw-recurrent-vae draw-recurrent-vae Public

    DRAW-inspired recurrent VAE for MNIST image generation: RNN vs LSTM, with/without attention (Gregor et al., 2015)

    Python

  4. Bias-detection-L3-project Bias-detection-L3-project Public

    Bias detection on a Chest X-ray (NIH 14) dataset

    Jupyter Notebook

  5. climbing-coach-ai climbing-coach-ai Public

    A computer vision system to analyze rock climbing technique from video footage. The core goal is to quantify climbing-specific metrics (e.g., hip-to-wall proximity, force vectors) and generate acti…

    Python

  6. pollinator-detection-challenge pollinator-detection-challenge Public

    Pollinator detection ML challenge (M1 AI, Paris-Saclay): designed and solved end-to-end, with dataset, baseline and scoring on Codabench

    Python 1