Headline

ML Engineer | NLP, Multimodal Learning and Model Interpretability | AAAI 2026 LM4UC Workshop and CVPR 2026 TRUE-V Workshop | IIT Madras, Vizuara AI Labs

About

I build systems at the intersection of NLP, multimodal learning, and model interpretability, with a focus on making AI more reliable in production and research settings. My research has been accepted at the AAAI 2026 LM4UC Workshop (fine-tuning NLLB-200 for Romanized Gujarati-English code-mixed translation with a self-constructed 30K parallel corpus, achieving 1.5 to 2x BLEU improvement over Google Translate) and the CVPR 2026 TRUE-V Workshop (geometric encoding of demographic attributes in ResNet-50 and ViT-Base). I also co-authored an arXiv preprint on NanoVLM, a parameter-efficient vision-language model, in collaboration with Vizuara AI Labs. On the engineering side, I have designed and deployed multimodal RAG systems handling real user traffic at ProCohat Technologies, built PPO-based RLHF pipelines for MusicGen at IIT Madras under Dr. Chandrasekhar, and conducted mechanistic interpretability research on deep vision models at Vizuara AI Labs. I am actively looking for ML Engineer, Research Engineer, and Applied Scientist roles in Bangalore. Open to connecting with researchers, engineers, and hiring managers working on hard model-level problems.

Location

Patna

AI stack

claude code

cursor

langchain

langgraph

huggingface

Supabase

Work experience

Software Engineer at Areta360 Technologies Private Limited, Jul 2026 - Present · 2 mos, Bengaluru

Engineer Intern at Areta360 Technologies Private Limited, Apr 2026 - Present · 5 mos, On-site

AI Research Intern at Vizuara Technologies Private Limited, Jan 2026 - Mar 2026 · 3 mos, Pune City, Work accepted at the CVPR 2026 TRUE-V Workshop. Investigated implicit demographic encoding in vision models using a multi-scale interpretability pipeline combining linear probing, filter-level correlation analysis, and representation geometry across ResNet and ViT architectures. Identified distributed subspace encoding of sensitive attributes and performed targeted filter ablations, reducing gender probing accuracy from 85% to 43% while preserving task performance and revealing causal bottlenecks.

AI Engineer Intern at FunnDynamix, Sep 2025 - Jan 2026 · 5 mos, Mumbai

Artificial Intelligence Intern at Wadhwani School of Data Science and AI, IIT Madras, May 2025 - Jul 2025 · 3 mos, Developed an ambient soundscape music generation pipeline using MusicGen, training on a curated 10-hour dataset and building a web platform to collect structured human preference feedback. Applied Reinforcement Learning from Human Feedback (RLHF) to align generated audio with human aesthetic and perceptual preferences using reward modeling and preference-based optimization.

AI Research Intern at Indian Institute of Technology, Guwahati, Nov 2024 - Mar 2025 · 5 mos, Guwahati, Assam, India

Education

Bachelor of Technology - BTech, Computer Science, Indian institute of information technology Nagpur

Public links

github: https://github.com/mansh7763

leetcode: https://leetcode.com/mansh__anand