Profile
Engineer with a strong foundation in software engineering and applied AI ,
including experience in developing:
-
Web Applications:
- Radiology labs management software
- 20+ cognitive games for elderly
- Distributed healthcare data platform for AI training
- Contract drafting software for lawyers
-
Applied AI systems:
- Pitch deck pre-screening tool for investors (Agentic AI)
- Conversational chatbot for pre-screening leads for real estate agents (Agentic AI)
- Speech-to-text software for lawyers in Hindi
- Radiology AI training systems using 2D, 3D CNNs and Vision Transformers
Checkout more details about these projects in my work experience or projects section.
Skills
- Web Development: React, Typescript, Next, FastAPI, Node.js, PostgreSQL
- Machine Learning: Speech to text, NLP
- Agentic AI: RAG, Chatbots, Fine-tuning, Evaluation, Observability
- Cloud & DevOps: AWS (EC2, S3, RDS, Route53, Cloudfront, ECR), Docker, Terraform
- Open Source Contributor: Stagehand (AI Browser Automation)
Education
B.Sc. in Data Science and Artificial Intelligence
Nanyang Technological University, Singapore (2018 - 2022)
President of IEEE NTU Student Branch (AY 2020-21): Led a team of 40+ members to organize 5 technical workshops, speaker events and a hackathon with 700+ participants, sponsored by SAP, HP & Govtech.
Artificial Intelligence Program
University of Science and Technology, Hefei, China (May-Jun 2019)
Experience
Research Assistant (Software Engineering)
Dementia Research Centre, NTU Singapore from Oct 2024 - Present
- Developed a full-stack web application for cognitive training including 20+ cognitive games using React, FastAPI & PostgreSQL. Checkout this blog for more design details!
- Applied CI/CD pipeline using GitHub Actions to automate full-stack deployment on AWS S3 & EC2.
- Optimized ML classification accuracy of cognitive impairment by ~4% using derived latent variables in tree-based models.
Software Engineer
Multiwave Innovation, Jaipur, India from May 2023 - Sep 2024
- Built a speech to text dictation webapp for drafting contracts using React, FastAPI & PostgreSQL used by over 150+ MAU (Monthly Active Users) with average weekday daily usage of 47 minutes.
- Prototyped an in-house React toolkit for real-time transcription with sub-500ms latency across multilingual inputs. Toolkit did the heavy lifting of connecting to speech-to-text model, Voice Activity Detection model & speaker diarization model via WebSockets by just entering the model endpoints. Tested the toolkit with Azure Speech and custom VAD, Diarization models from pyannote.
- Enhanced ASR model performance by 6% through LLM-based transcription error correction during post-processing using FlowiseAI
Co-founder
Diagnokare, Bangalore, India from May 2022 - May 2023
- Developed a federated learning platform for radiology AI researchers to simplify data access for 6 enterprise clients using React, FastAPI and PostgreSQL.
- Created a radiology lab management software managing over 1M+ records.
- Optimized CT & MRI scan viewer latency by 36% by shifting to Cloudfront CDN (Content Delivery Network).
- Deployed the platform on AWS using EC2, S3, RDS for enterprise usage.
Analyst (Internship)
Noviscient <> SGInnovate, Singapore from Dec 2020 - May 2021
- Prototyped a webapp for AI-driven portfolio management using Voilia on Jupyter Notebooks.
- Developed probabilistic machine learning algorithms to predict fund returns using PyMC3.
- Generated financial synthetic datasets using multiple deep learning based models for backtesting.
Data Scientist (Internship)
Singapore Airlines, Singapore from Jun 2020 - Jul 2020
- Trained and deployed NLP-based machine learning models to classify consumer feedback to appropriate department saving manpower cost by 15 hours per week.
- Applied feature extraction techniques like TF-IDF, Glove & BERT to improve classification accuracy on ML models like Naive Bayes, Random Forest, XGBoost, SVMs.
- Reduced WER on Singlish by ~6% on speech to text tasks using feature engineering on DeepSpeech 2 on 550 hours of speech data totalling to ~63 GB.