MohamedHamlil

• A neural network learning to play Flappy Bird from raw pixels 0.4 to 12.65 pipes, from one colour channel after-hours
• Sorting algorithms rendered in Blender, one swap per frame after-hours
• A Battleship bot working out where the ships are 74.3% of games, 54.8 shots to clear University coursework
• Cutting the multiplications a polynomial product needs 25 multiplications to 14, and the rank proved to be 13 Rank of bilinear maps
• A SAT solver that turns logic into algebra over GF(2) 500 instances, zero wrong verdicts after-hours
• What the models actually put in the message 0.39 of messages name an action, down to 0.00 AI and strategic pricing
• What actually moves together in 24 000 fragrances University coursework
• Predicting how a perfume will be rated, four models compared best AUC 0.693, all four between 0.65 and 0.69 University coursework
• An imitating agent meeting a cooperator, a defector and a coin flip 8th of 8, behind a coin flip AI and strategic pricing
• Two models keeping whatever regime they were handed 0.21 cooperating out of a defecting opening, 0.47 with a message AI and strategic pricing
• What French income tax actually costs, marginal against average inflection at €59,800 after-hours
• An FFT written from scratch, checked against NumPy agreeing with NumPy to 1e-13 after-hours
• Sixteen Prolog strategies played against each other 7th and 8th of 16 AI and strategic pricing
• How an agent that copies what it sees changes its mind reciprocity to zero, Tit-for-Tat holds at one AI and strategic pricing

Each picture opens the project it came from, where the numbers behind it name the file that produced them. Reachable at mohamed.hamlil@etu.univ-grenoble-alpes.fr, on LinkedIn, and on GitHub. More ways and what I am looking for.