Tejas Bana

Avatar

I am a Machine Learning Engineer at Albert Invent, where I build reliable, scalable, and real-time agentic AI systems to accelerate scientific R&D. My work focuses on creating LLM agents through multi-agent architectures, relational Retrieval-Augmented Generation (RAG), and prompting techniques.

Apart from my current work, I am interested in representation learning and diffusion models. My background includes a Master's from Carnegie Mellon University and a portfolio of projects spanning generative models, computer vision, and privacy-preserving ML.


LLM Agents & Applied AI


Generative Models & Representation Learning

GANs for Architectural Floorplan Generation

ML Intern at Uniworks Designs

GANsPix2PixGCPFlask

Automated the generation of apartment floorplans and furniture layouts using a two-stage GAN pipeline (House-GAN and ArchiGAN). Deployed the model as a service on GCP, reducing initial design time for architects by up to 25%.

Generative Camouflage for the Indian Army

R&D ML Intern at College of Military Engineering

Computer VisionGenerative ModelsMask R-CNNK-Means

Developed software to generate pixelated, terrain-specific camouflage patterns. The system uses satellite imagery, segments terrain features with Mask R-CNN, and applies a generative algorithm to create 5K resolution patterns in under 60 seconds, significantly reducing vehicle visibility. Received "Best Intern" award for this work.

Learning-Based Image Synthesis (16-726)

Carnegie Mellon University

GANsNeural Style TransferComputer Vision

A deep dive into generative vision through hands-on projects:

Generative Vision Seminar (16-895)

Carnegie Mellon University

Diffusion ModelsGenerative ModelsSeminar

A seminar course exploring the nuances of modern generative models, from understanding their success to critiquing their capabilities for future research. Covered topics like Diffusion Models, VAEs, and Flow-based models.

Course Page

Siamese Networks for Fashion Similarity

R&D ML Intern at IIT Bombay

Representation LearningPyTorchSiamese NetworksTriplet Loss

Built a deep learning system to find visually similar clothing items from a massive dataset. Implemented a Siamese network with triplet loss and multi-class negative sampling to learn powerful image representations. Used Annoy for efficient, real-time nearest-neighbor search.


Publications

  • Robust Recovery of Adversarial Samples
    Tejas Bana, Jatan Loya, Siddhant Kulkarni
    ICML-2021 Workshop on AML
    Paper

  • Privacy-Preserving Keystroke Analysis using Fully Homomorphic Encryption & Differential Privacy
    Jatan Loya, Tejas Bana
    International Conference on Cyberworlds (CW 2021)
    Paper
  • Vit - Inception - GAN for Image Colourizing
    Tejas Bana, Jatan Loya, Siddhant Kulkarni
    The Conference on Machine Learning, Optimization and Data science (LOD)
    Paper


Relevant Coursework

  • On-Device Machine Learning (11-767): Built efficient ML systems via quantization, pruning, and neural architecture search.
  • Large Language Models (11-667): Studied methods and applications of modern LLMs.
  • Advanced NLP (11-711): Audited course covering advanced topics in Natural Language Processing.
  • Neural Signal Processing (18-698): Learned neural decoding, firing rate estimation, and spike sorting.
  • Computational Models of Neural Systems (15-883): Explored computational approaches to understanding neural function.
  • Introduction to Machine Learning (18-661): Core principles and algorithms of machine learning.