welcome to mohan's home on the internet

I work @ Intelligent Vehicles Lab
I'm also working towards my Ph.D in Deep Learning and Intelligence Simulation @ Munich University of Applied Sciences

current research objective:

making AI smarter by understanding and modeling human behaviour in real-time. I work on pose estimation, pedestrian modeling, and integrating these with autonomous systems—particularly in driving. The goal is simple: accelerate general AI that can see the world like we do, understand it, and act accordingly.

other stuff i like:

simulation hypothesis, human-computer interactiions, game development, neuroscience and game theory.

outside research
i do a lot a weightlifting. sometimes post here INSTAGRAM
i consolidate my thoughts here X

here is some enterpise work:

HM logo

2023*

my lab IVL is at Munich University of Applied Sciences. I do research there.

altada logo

2021 - 2022

was doing document intelligence for a FinTech startup, computer vision and natural language processing

stats logo

2022

performed rugby statistics and data analysis work at StatsPerform.

UL logo

2020 - 2022

did a masters degree at the University of Limerick. mainly to learn business in depth after my first startup experience. Alongside this, i did a lot of data science and machine learning during this time

PES logo

2015 - 2019

my time at the PES University was full of hackathons, electronics, physics and sports. during that time i did an internshipt with the crucible of reseach and innovation during their satellite (PISAT) launch with IRSO.

makervillage logo

2016 - 2018

started my first company with my friends to build a hardware-ml parking solution.

publications:

Walk-the-Talk: llm driven pedestrian motion generation

the field of autonomous driving, a key challenge is the "reality gap": transferring knowledge gained in simulation to real-world settings. Despite various approaches to mitigate this gap, there's a notable absence of solutions targeting agent behavior generation which are crucial for mimicking spontaneous, erratic, and realistic actions of traffic participants. Recent advancements in Generative AI have enabled the representation of human activities in semantic space and generate real human motion from textual descriptions. Despite current limitations such as modality constraints, motion sequence length, resource demands, and data specificity, there's an opportunity to innovate and use these techniques in the intelligent vehicles domain. We propose Walk-the-Talk, a motion generator utilizing Large Language Models (LLMs) to produce reliable pedestrian motions for high-fidelity simulators like CARLA. Thus, we contribute to autonomous driving simulations by aiming to scale realistic, diverse long-tail agent motion data – currently a gap in training datasets. We employ Motion Capture (MoCap) techniques to develop the Walk-the-Talk dataset, which illustrates a broad spectrum of pedestrian behaviors in street-crossing scenarios, ranging from standard walking patterns to extreme behaviors such as drunk walking and near-crash incidents. By utilizing this new dataset within a LLM, we facilitate the creation of realistic pedestrian motion sequences, a capability previously unattainable (cf. Figure 1). Additionally, our findings demonstrate that leveraging the Walk-the-Talk dataset enhances cross-domain generalization and significantly improves the Fréchet Inception Distance (FID) score by approximately 15% on the HumanML3D dataset.

Mohan Ramesh, Fabian B. Flohr 2024
Walk-the-Talk

more later...

unorganized brain dump (don't read):

nov 2023 Building an NFT startup: Just a brain dump
aug 2022 magical window function
jan 2022 do artifacts have politics?
jan 2022 squats and startups
nov 2021 what is the purpose of life
oct 2021 what does technology do to a nation?
may 2021 what it takes to be a leader
mar 2021 simulation hypothesis
dec 2020 can businesses compete on data?
nov 2020 the analytics challenge for an org.
nov 2020 decision making is an art!
no idea housing price prediction
no idea simple tweet classification
no idea topic modeling
no idea whatwasithinking