LinguaSense
Multilingual speech processing under code-switching and background noise.
At 10 dB SNR, WER fell from 31.45% to 9.20% with DeepFilterNet (Table III).
I’m Ronit Reddy, a final-year Computer Science and Engineering student at Woxsen University, working on multilingual speech, regulatory question answering, visual assistance and EEG classification.
I’m interested in how AI systems ground their answers in evidence, how we evaluate their reliability, and how people understand and use their outputs.
From evidence-grounded language systems to evaluation on unseen participants.
Multilingual speech processing under code-switching and background noise.
At 10 dB SNR, WER fell from 31.45% to 9.20% with DeepFilterNet (Table III).
Evidence-grounded question answering over pharmaceutical regulatory documents.
Held-out descriptive evaluation: faithfulness 0.960 on 25 questions.
A voice-controlled visual-assistance prototype for blind and low-vision users.
Placed fourth among 1,800+ teams at the NextGen AI 2026 National Hackathon, held in collaboration with Google Gemini.
A four-stream temporal CNN for classifying rest and imagined movements in previously unseen participants.
A 19.94M-parameter variant achieved 70.23% mean accuracy and 59.68% macro-F1 with 76% fewer parameters than the reference.
A collaborative adversarial-testing prototype, with a documented open backend contribution.
Project details →A watch-party app combining WebRTC screen sharing with synchronized playback.
Project details →A medical-record management prototype built around ownership, roles and access policies.
Project details →AI Engineer Intern
Building an enterprise lead-generation and qualification assistant within a four-person team.
Biomedical Signals and AI Research Intern
Studying EEG motor-imagery classification with subject-independent evaluation and reproducibility.
Related project →Software Development Engineer Intern
Contributed to natural-language-to-SQL functionality for an enterprise AI-agent initiative.
Placed fourth among 1,800+ teams at the NextGen AI 2026 National Hackathon, held in collaboration with Google Gemini.
Quantum Machine Learning Models for Acoustic Sensor Metrology.
One of five Woxsen University students on the 2025 exchange.
During my exchange semester in Malaysia, I became more comfortable approaching people without having everything figured out first. Conversations with professors led to work that continued after I returned home. Back at Woxsen, I also help juniors when they get stuck and participate in club workshops. Having often needed someone else’s time and patience, I try to offer the same.
For research conversations, collaboration and opportunities.