Apoorv Agnihotri

AI Researcher

Tübingen, Baden-Württemberg, Germany3 yrs 9 mos experience

AI EnabledAI ML Practitioner

Key Highlights

4+ years of experience in machine learning research.
Developed AI solutions impacting healthcare and education.
Expert in voice cloning and generative models.

Stackforce AI infers this person is a Machine Learning Specialist with a focus on AI applications in healthcare and education.

Contact

Skills

Other Skills

Active LearningAlgorithmsApplied Machine LearningArtificial Intelligence (AI)Bayesian OptimizationC++CUDAComputer VisionData ScienceData StructuresData VisualizationDeep LearningGenerative AIGenerative Adversarial Networks (GANs)Generative Research

About

I’m a machine learning researcher and builder with 4+ years of experience applying deep learning to real-world problems across speech, healthcare, and education. From deploying AI models in rural India to developing neuromorphic computing prototypes at Mercedes-Benz, my work lies at the intersection of scalable systems and meaningful impact. At Rephrase (acquired by Adobe), I worked on voice cloning and speech safety using VAEs and GANs. At Wadhwani AI (Gates Foundation–backed), I helped deploy ML for medical document processing and pest monitoring in agriculture. At ArtPark, I explored how LLMs (via RAG) can assist community health workers on the field. Currently, I’m finishing my Master's in Machine Learning at the University of Tübingen, where I’m researching on using LLMs to help math teachers in Germany incorporate pedagogical best practices. I'm particularly drawn to roles where I can build and scale AI products that are elegant in design, robust in deployment, and useful to people who need them most.

Experience

Mercedes-benz ag

Working Student - AI Research

Aug 2024 – May 2025 · 9 mos · Böblingen, Baden-Württemberg, Germany · Remote

Artpark

Machine Learning Specialist Consultant

Jul 2023 – Sep 2023 · 2 mos · Bangalore · Hybrid

Built a prototype using RAGs (retrieval augmented generation) on LLMs to improve knowledge dissemination to Auxiliary Nursing Midwifery to help in quicker help with pregnancy care on the field.

Rephrase.ai (acquired by adobe)

Deep Learning Researcher

Apr 2022 – Jun 2023 · 1 yr 2 mos · Bengaluru, Karnataka, India

Speech Synthesis:
Used Variational Auto-encoders (VAEs) and Generative Adversarial Networks (GANs) to work on generative speech, creating a proof of concept for voice cloning with few seconds of target speaker's voice samples.
Used off-the-self speech content encoders (hubert-soft) to flag problematic speech on production systems. Further, my explorations in speech encoders helped us improve our lip-sync model.
Machine Learning Engineering:
Improved team's experimental throughput by integrating ML experimenting frameworks like pytorch-lightning.

Wadhwani ai

2 roles

Associate ML Scientist I

May 2021 – Apr 2022 · 11 mos · Mumbai, Maharashtra, India

Research Fellow

Jun 2020 – May 2021 · 11 mos · Mumbai, Maharashtra, India

Led the development of a research prototype for scanning and interpreting medical docs using template matching and classical computer vision.
For our mobile solution, implemented and deployed SOTA Deep Learning techniques for object detection.
Experimented with sample rejection of adversarial inputs in the wild. (NeurIPS 2022 workshop, ICLR 2023)
Coauthored a white paper detailing potential venues for AI disruption in Indian agriculture.
Represented Wadhwani AI at Google's AI for Agriculture Hackathon and won a cash reward of 1 million rupees.

Indian institute of technology gandhinagar

2 roles

Teaching Assistant

Feb 2020 – Mar 2020 · 1 mo · Gandhinagar, Gujarat, India

Created slides and exercises in ML, targeted to help professors from different fields, use ML. I further helped with the administrative tasks during the Technical Education Quality Improvement Programme.

Summer Research Intern

May 2019 – Jul 2019 · 2 mos · Gandhinagar

wrote an expository article on Bayesian Optimization. (Distill 2020)
created an open‑source repository (polire) for spatial interpolation and sensor placement. (Poster at SenSys 2020)
experimented using active learning for sensor placement recommendation to reduce costs. (CoDS COMAD 2020, ICML Workshop 2020

Nvidia

Accelerated HPC and Machine Learning Intern

May 2018 – Jul 2018 · 2 mos · Bangalore, India

contributed GPU‑optimized Kalman filters to rapids.ai, an open‑sourced data science framework for ML algorithms.
created a library to generate multivariate Gaussian random numbers on GPU, subsequently reused by various projects