Ayush Kumar
CSE '27 · SRM Chennai
Command Palette
Search for a command to run...
Software Engineer | AI Engineer | Product Engineer | Samsung PRISM Research Intern | 2× National Hackathon Winner | Building AI Agents, Developer Tools & Intelligent Products
Experiences
100K+
Records
78 - 97%
Accuracy
Mar-Aug
Term
3
Domains
- •Predictive Modeling: Built models for delinquency risk, marketing performance and sales forecasting on 100K+ real-world business records
- •Accuracy: Improved prediction performance from 78% to 97% through feature engineering and experimentation
- •Decision Support: Developed workflows that turned raw business data into actionable insights across multiple use cases
- •Evaluation: Measured model performance with quantitative metrics and presented findings to stakeholders
10,000+
Students
Backend
Role
Sep-Apr
Term
Data
Focus
- •Workflows: Built digital workflows that streamlined resume management and placement processes for students
- •Automation: Reduced manual administrative effort by automating document generation and tracking
- •Data Models: Designed scalable data models for user profiles, resumes and application records
- •Scale: Contributed to a platform supporting placement workflows for 10,000+ students annually
- •Collaboration: Worked closely with product, design and engineering teams to deliver user-focused solutions
First Author
Role
BiLSTM
Model
SHAP
XAI
SMM4H
Venue
- •Paper: Time-to-Stress Prediction in Precision Agriculture, a Dual-Head BiLSTM approach with SHAP-based temporal attention
- •Approach: Designed a dual-head BiLSTM with SHAP-based temporal attention for interpretable time-to-stress forecasting
- •Service: Selected as a Reviewer for SMM4H-HeaRD 2026
Projects
Play Video
Samsung PRISM — Worklet 8
ML intelligence dashboard serving three trained models in real time — loan delinquency risk (RandomForest, AUC 0.72), campaign performance (CatBoost · LightGBM · Ridge) and sell-out forecasting (XGBoost). Models are exported to ONNX in CI with drift-verified builds and run live on Vercel.
Play Video
Nomi
An AI companion that lives beside your browser — a floating glass avatar that learns your habits, builds a daily memory of your work and answers questions like "when did I last study graphs?". Local-first: on-device inference with transformers.js, encrypted storage, no cloud recording.
Play Video
LoopJob
A personal always-on agent that monitors company career portals, discovers matching jobs across the indexed job market and semantically matches them against your filters — then emails you new openings, deduplicated, with the reason each one matched.
Play Video
Rate Limiter
Distributed API rate-limiting service on Spring Boot 4 and Redis — token-bucket and sliding-window strategies backed by atomic Redis operations, so limits stay consistent across service instances.
GitHub Activity
Loading GitHub contribution activity
Calendar heatmap showing daily GitHub contribution counts for Ayush277 over the last year. Scroll horizontally to inspect all weeks.
LeetCode Activity
Loading LeetCode activitySkills & Technologies
Python
C++
JavaScript
HTML
CSS
React
FastAPI
MySQL
MongoDB
TensorFlow
Pandas
NumPy
scikit-learn
Git
GitHub
Docker
Excel
PowerPoint
LinuxLeadership & Achievements
2× National Hackathon Winner
1st place at IntelliHack SRM and AIR 3 at the BITS Hyderabad All-India Hackathon.
Student Technology Organization Lead
Led a 100+ member organization for two years — recruitment, operations, event planning and cross-functional teams.
Published Researcher & Reviewer
Authored a Dual-Head BiLSTM with SHAP-based temporal attention. Selected as a Reviewer for SMM4H-HeaRD 2026.
Competitive Programming
Solved 200+ algorithmic problems across LeetCode and GeeksforGeeks.
UI/UX & Graphic Design Freelancer
Delivered branding assets, design systems and digital experiences for startups and student organizations.