I work across machine learning, data-driven systems, and full-stack development with a focus on measurable outcomes, elegant interfaces, and production-minded execution.
ML Systems
Applied AI workflows.
Full Stack
React, APIs, UX.
Now Shipping
Live project work.
About
I build AI and autonomy systems with enough product thinking to make the work usable, legible, and ready for real workflows.
Current research
Building a physics-grounded Isaac Sim smart-assembly testbed with a TurtleBot3 transporter, four Franka arms, lidar navigation, and ROS 2 coordination for manufacturing research.
Applied ML impact
First-author work accepted at ESCS'26 / Springer, pairing a six-class YOLO11 inspection model with live annotated detections, pass/review verdicts, and exportable results.
40%
accuracy lift
DRDO audio anomaly detection
0.99
mAP50
PCB defect detection capstone
ROS 2
robotics stack
Isaac Sim smart-manufacturing research
4
AI agents
healthcare decision-support platform
Education
CSUF
M.S. Computer Science (Aug 2024 – Aug 2026)
GPA 3.78
JNTUH
B.Tech Computer Science & Engineering
GPA 7.87
Selected Work
A focused set of shipped work across applied AI, full-stack tools, and user-facing engineering.
GitHub profileSix-class PCB inspection system using YOLO11l and RT-DETR-L benchmarks, delivering 98.3% precision, 98.8% recall, and 0.99 mAP50 in a live Hugging Face inspection app.
Four-agent healthcare decision-support platform with orchestration, retrieval, recommendations, explainability, safety guardrails, audit logging, and a Dockerized REST API.
Production-style transaction fraud scoring on the IEEE-CIS dataset: XGBoost inference, cost-based thresholding, Tree SHAP explainability, Evidently drift monitoring, FastAPI service, and a Streamlit workbench.
NLP-driven trading simulation combining BERT-scored financial news, Yahoo Finance data, confidence-weighted trade decisions, and historical backtesting.
TensorFlow/Keras autoencoder pipeline for industrial fan audio anomaly detection using Librosa log-mel spectrogram features and reproducible evaluation workflows.
Deep-learning image classifier using Inception-ResNet-V2 across 120+ dog breeds with 90%+ accuracy, GPU-accelerated training, and OpenCV preprocessing on 10,000+ images.
Student-facing academic advising platform with a responsive React interface, Firebase-backed real-time data integration, and usability-focused workflows.
Flask image-classification service with batch uploads, confidence scoring, health and metrics endpoints, image-hash caching, retry handling, pytest, and CI.
Cafe discovery app with geolocation, Google Maps & Places APIs, and a card-based UI for nearby place discovery.
Open Source
Pull requests across open-source ML and developer tooling repos, including merged work on falsify-inspect and licinexus-mcp.
Live Activity
Pulled from recent public GitHub events plus direct repository commit checks, including this portfolio.
Capabilities
Resume
Download my latest resume for experience, education, technical skills, and project outcomes.
Download Resume