Built an end-to-end ML pipeline predicting customer review sentiments from raw e-commerce data.
Built a custom C page-based storage engine with WAL and 2PL for a high-throughput ticketing server.
Stellar Classification. Built a stacked ensemble with balanced-accuracy optimization.
Architected normalized relational databases, REST APIs, and multi-threaded TCP servers.
Overview_
What I Build
I engineer intelligent systems, mitigate algorithmic bias, and build high-performance deep learning architectures. My work focuses on scalable machine learning infrastructure and advanced natural language processing.
Projects
Hidden Layer 3_
Event Ticketing Platform
- Engineered a multi-threaded TCP server in C utilizing bounded thread pools and select() multiplexing.
- Built a custom storage engine featuring a Write-Ahead Log and Two-Phase Locking (2PL).
Hidden Layer 3_
Traffic Sentinel AI
- Trained a 3-stage YOLOv11 cascade pipeline on 17,670 images for real-time traffic detection (mAP50 0.936).
- Boosted inference throughput by 1.8x by converting PyTorch models to ONNX Runtime.
Hidden Layer 4_
Customer Review Prediction
- Built an end-to-end ML pipeline to predict review sentiments on 100K+ e-commerce orders.
- Achieved 0.829 F1 with a Random Forest and derived 5 actionable business insights.
Hidden Layer 4_
Air-Cargo Database Engine
- Designed a normalized 5-table relational schema and built a REST API backend.
- Ensured zero data corruption with ACID transactions and secured endpoints via JWT Auth.
Experience
Hidden Layer 6_
NLP Research Intern
@ IIIT Bangalore
- Architected an autonomous LLM data-extraction pipeline with AI-as-a-Judge validation over the OpenAI API, converting 60,000+ words of unstructured anthropological text into 4,785 verified cultural rules of thumb across 9 Indian states, achieving 95%+ extraction precision.
- Spearheaded development of a cross-cultural evaluation benchmark spanning 29 Indian regional cultures to systematically measure Western bias in LLMs, targeting a top-tier peer-reviewed NLP venue.
Achievements
Hidden Layer 8_
Datathon Finalist
Data Driven Policy Innovation (IIIT Bangalore, Synergy '25): advanced to finals among 333 registered teams.
Hidden Layer 8_
Top 16% Global Rank
Kaggle Stellar Classification. Built a robust 9-model cross-validated meta-stacker with a defensive consensus architecture.
Output_