AI/ML Researcher • Open‑source Developer • AI Ethics & Bias Mitigation • Advocate for Centrally‑Localized (On‑Device/Self‑Hosted) LLMs • Public Speaker

I build practical, privacy‑respecting AI systems that run locally and make a real impact. I care deeply about accessibility, model efficiency, and responsible AI. My work spans speech recognition, LLM fine‑tuning (LoRA), generative imaging, fuzzy logic control, and data tooling—always with a bias toward clear UX and simple, reproducible installs.

• Project: Subtitles for Visual Impairment Assistance — local, lightweight live speech transcription
• Why I built it: After being diagnosed with ear injuries and dealing with intermittent tinnitus, I relied on Chrome’s live captions. When I switched to Firefox and lost that feature, I built a locally hosted, open‑source alternative with a simple install.
• Highlights:
  • Fully local; total footprint ~150 MB (deps + model)
  • Two modes:
    • Live: big, highlighted words in real time (TikTok/Reels‑style)
    • Comprehensive: continuous paragraph view for long talks where full context matters
  • Designed for accessibility, privacy, and low overhead
• Repo: https://github.com/SatyamSaxena1/Subtitles-for-Visual-Impairment-Assistance-

• Intelligent Joining Speed on Highways (Fuzzy Logic)

  • A fuzzy‑logic system that ingests lane condition, oncoming traffic, lane density, vehicle distance, etc., and recommends near‑instant optimal merge speed using >100 rules with Gaussian/logistic/singleton membership functions.
  • Repo: https://github.com/SatyamSaxena1/fuzzy-logic-highway-proj

• Reddit Saved‑Posts → Mind‑Map/Notion Board

  • A WIP tool to analyze your exported saved posts (starting with Reddit), cluster/group them, and resurface forgotten ideas in a mood‑board style workflow.
  • Repo: https://github.com/SatyamSaxena1/reddit-scraper-for-mind-map-project

• AI/ML: Speech recognition and transcription, NLP/LLMs (prompting + LoRA fine‑tuning), fuzzy logic systems, recommender systems, classical ML
• Generative AI: Stable Diffusion (SDXL) pipelines, Kohya_ss training, BLIP, safetensors
• Optimization & Efficiency: LoRA, Adafactor, BF16 precision, small‑footprint local deployments
• Data: CSV ETL pipelines, structured data modeling, data segmentation, caption augmentation
• Tooling & Frameworks: PyTorch, Hugging Face Diffusers, OpenCV, Tableau
• Languages: Python, C++, R
• Platforms & Hardware: Anaconda, NVIDIA T4 / RTX 3080 Ti; Google Colab, AWS
• Communication: Public speaking, clear UX thinking, documentation, teamwork/leadership

• MSc, Artificial Intelligence — Asia Pacific University of Technology and Innovation (APU/APIIT) & De Montfort University (Dual Degree), 2023–2024
  • Activities: Software Engineering Project Showcase, Open Source Contribution, Technology & Innovation Society, Programming Club
• B.Tech, Computer Science & Engineering — G.D. Goenka University, 2022
  • Specialization in AI & Machine Learning

• Ethical AI and bias mitigation baked into the lifecycle
• Privacy‑first, centrally‑localized/on‑device LLMs
• Sustainable, efficient systems that are easy to install and use
• Sharing knowledge through open source and public speaking

Email only: [email protected]