ETH Zurich
MSc in Data Science. Coursework includes Computer Vision, 3D Vision, Probabilistic AI, NLP, Deep Learning, and Computational Statistics.
ETH Zurich - MSc Data Science
Adhithya LaxmanR G
Third-semester MSc Data Science student at ETH Zurich, working across computer vision, 3D reconstruction, and production-minded AI engineering.
$ cat focus.txt
AI engineering, ML systems, computer vision, 3D reconstruction
$ ./open_to_work --roles swe,ml-systems,cv
OK available for research-driven engineering conversations
About
Building practical ML systems with research depth.
I am a third-semester MSc Data Science student at ETH Zurich. My work sits at the intersection of computer vision, 3D reconstruction, retrieval, and ML systems, with experience in research pipelines, edge deployment, and data-heavy engineering.
Before ETH, I completed a B.E. in Computer Science at Coimbatore Institute of Technology and a Diploma in Data Science from IIT Madras in parallel.
Languages
Python
C++
Java
C
Shell Script
SQL
ML & Data
PyTorch
TensorFlow
ONNX
Scikit-learn
NumPy
Pandas
Flask
Django
Systems & Cloud
Docker
Git
Linux
Kubernetes
AWS S3
GCP
Azure DevOps
Jenkins
Data Infrastructure
Spark
Hive
ClickHouse
PostgreSQL
MongoDB
Grafana
Focus Areas
Computer Vision
3D Reconstruction
Deep Learning
RAG
Edge AI
NLP
Transformers
Multimodal VLMs
Distributed Systems
MLOps
Education
Academic background
Coimbatore Institute of Technology
B.E. Computer Science. CGPA 9.57, top 2% of the class.
IIT Madras
Diploma in Data Science and Applications, completed alongside undergraduate studies.
Experience
Research and engineering roles
AI Engineering
AI Engineering @ Robinreal AI
Working on applied AI systems for real-estate workflows, with a focus on retrieval, agent tooling, backend reliability, and evaluation for production-facing ML features.
Research
ETH Zurich - Computer Vision and Geometry Group
Improved long-horizon indoor 3D reconstruction for in-the-wild videos by upgrading the HouseTour pipeline with MegaLoc + AIR-HLoc retrieval, grid-anchored RoMa v2 tracks, and COLMAP mapping.
Recovered 13 of 30 previously failed HouseTour sequences while retaining all prior successes; successful reconstructions reached 100% keyframe registration with strong multi-view track statistics.
CV / ML Systems
Software Engineer Intern @ Caterpillar Inc.
Developed a monocular depth-estimation pipeline for autonomous truck perception, training on 50k+ images and optimizing inference for edge AI hardware.
DevOps
Software Engineer Intern @ Akamai Technologies
Built Jenkins-based ETL pipelines for large-scale Hive data, integrated ClickHouse and Grafana monitoring, and modernized legacy data pipeline components.
Research
Research Intern @ Samsung R&D Institute PRISM
Led a team of five building Conditional GAN models for MNIST-based video sequence generation, reaching 0.97 SSIM and receiving an Excellence Certificate.
Deep Learning
Research Intern @ CSIR-CSIO
Explored neural-network models for solar panel fault detection and compared optimizers for fault classification performance.
Projects
Selected work
Relational Reasoning in Vision-Language Models
Analyzing cross-modal attention and shortcut learning in VLMs using synthetic datasets, ablations, activation patching, and probing.
View projectFederated IDS with Deep Belief Networks
Led a team of five building a privacy-preserving WiFi intrusion-detection system with Go/Python gRPC training and 98% accuracy.
View projectDoc-Chat
Built a PDF chatbot with ingestion, semantic chunking, vector indexing, grounded Q&A, summarization, and quiz generation.
View projectBlood Connect
Created a blood donation management system with donor records, request workflows, and a PostgreSQL-backed web interface.
View projectLyric Lullaby
Developed a music streaming app with authentication, playlist management, search, and relational data modeling.
View projectHandwriting Letter Classification
Built a feed-forward neural network for handwritten character recognition, achieving over 90% accuracy on EMNIST.
View projectRecognition
Selected achievements
Springer 2025 accepted paper
FEDDBN-IDS: Federated Deep Belief Network for Intrusion Detection on Heterogeneous WiFi Networks.
LeetCode Knight
Top 2.5% globally with a 2020+ rating, reflecting sustained algorithmic problem-solving practice.
TheAlgorithms/Python contributor
Contributed neural-network algorithms and optimization techniques to a 180k+ star open-source repository.
SIH and Innovision placements
Winner at Internal Smart India Hackathon 2023 and 2nd place at Innovision24 among 48 teams.