- 🎓 B.Tech student, Artificial Intelligence & Machine Learning — CGPA 8.9, Bengaluru
- 🛠️ I build end-to-end systems, not notebooks — REST APIs, database-backed auth, and applied ML, shipped with tests and documented trade-offs
- 🤖 Recent work spans document intelligence (OCR + semantic search), secure session-based auth, agentic AI systems (RAG, MCP, LLM tool-calling), and geospatial/ML crime & risk analytics
- ☁️ Comfortable across the stack: Python/FastAPI, Node/Express, React, PostgreSQL, and cloud deployment (Google Cloud Run, Zoho Catalyst)
- 🚀 Currently looking for a Software Engineer internship
| Languages |
|
| Backend & APIs |
|
| Databases |
|
| AI / ML |
|
| Frontend |
|
| Security |
|
| Cloud & Tools |
|
Node.js · Express · PostgreSQL · Appwrite · bcrypt · Session Auth
- Two parallel implementations of the same auth/file-access contract — a hand-rolled Node/Express + PostgreSQL backend, and a managed Appwrite backend — to compare what a platform automates vs. what needs explicit configuration
- Server-side sessions with ID regeneration on login (fixation prevention), identical
404s for missing vs. not-yours resources to close an ID-enumeration side channel, and attempt-based account lockout - 15 automated tests + manual
curlverification of every cross-user isolation case - (Related earlier submission: FOSSEE — same system, built for the FOSSEE fellowship)
Python · FastAPI · Tesseract OCR · FAISS · Sentence Transformers · OpenCV
- Local, zero-cloud document intelligence pipeline: OCR + table extraction + regex NER, editable results, and semantic search across the whole document corpus
- 94.2% average OCR confidence, 88.7% table extraction accuracy, ~38ms semantic search latency, benchmarked on a 100-document corpus
- Human-in-the-loop correction pipeline that feeds back into the extraction model
FastAPI · React · XGBoost · SHAP · PostGIS · Zoho Catalyst
- Built for the Karnataka State Police Datathon 2026 — geospatial hotspot detection, criminal-network analysis (MO similarity clustering), and XGBoost + SHAP-explained weekly station risk prediction
- Validated against a synthetic dataset with 7 injected ground-truth patterns — risk model beat a naive baseline (MAE 5.00 vs. 6.66) and correctly recovered all 5 injected repeat-offender rings
- Deployed live on Zoho Catalyst with a Supabase/PostGIS backend
- (Earlier solo submission: Crime-Intelligence-Platform)
FastAPI · React · pgvector · PostgreSQL · RBAC
- Multi-tenant internal AI assistant combining RAG over company documents, a structured HR database, a deterministic business-rule engine, and role-based access control
- Every tenant-scoped query is filtered by
company_idfrom the JWT — never a client-supplied value — with pgvector-backed document retrieval scoped per tenant - Business rules (e.g. leave-balance checks) are computed deterministically in Python and handed to the LLM as ground truth so it can't hallucinate a number
Python · OpenCV · FastAPI · React Native scaffold
- Classical CV pipeline (CLAHE + dual-threshold blob detection + Canny edge detection) to detect and score pothole/crack severity from road images
- Turns a batch of images into a priority-ranked, cost-estimated repair queue for road-authority planners
- Explicitly documents what's a working demo vs. a production upgrade path (trained detector, real camera calibration, sourced cost data)
FastAPI · Node.js · MongoDB · React · Scikit-learn
- Privacy-first, explainable movie recommendation engine — diversity-mixed recommendations (40/30/20/10 blend + backfill), hidden-gem surfacing, and mood/time/group-aware filtering
- 17/17 recommendation-logic tests passing; verified service boot, routing, and schema validation across the full stack
Flask · SQLAlchemy · SQLite · Bootstrap 5
- Production-style HR Management System with real organizational hierarchy, multi-stage leave approvals, reversible archive/restore workflows, and a full audit trail
- Backend-enforced RBAC across 5 roles (Employee → Manager → HR → CEO → Admin) — authorization is never just a hidden UI button
- Includes an offline, rule-based "Dayflow Assistant" that answers HR questions from the live database with zero external API calls
| Project | Description |
|---|---|
| Secure-Login-System / FOSSEE | Dual-backend session auth + IDOR-safe file access (see above) |
| KSP-Hackathon-Deploy / Crime-Intelligence-Platform | Crime hotspot + risk-prediction platform (see above) |
| Degree | Institution | Score |
|---|---|---|
| B.Tech, Artificial Intelligence & Machine Learning | 2028 | CGPA 8.9 |
| 12th Grade (PUC) | CJC, Bengaluru — 2024 | 90.8% |
| 10th Grade (SSLC) | SEI, Bengaluru — 2022 | 97% |
Sustainable AI Intern — 1M1B (1 Million for 1 Billion) · Dec 2025 – Jan 2026 · Remote
- Built and debugged AI models for sustainability use cases, improving inference efficiency and reducing compute cost
- Automated data-preprocessing and training pipelines in a remote Agile team across the full SDLC
- NPTEL — Data Structures & Algorithms
- NPTEL — Introduction to Database Systems
- HackerRank — Python & C++ (Gold)
- Oracle — Java Certified Foundations Associate
- Oracle — Agentic AI Certified Foundations Associate
- Oracle — OCI 2025 Certified AI Foundations Associate
- Forage — Tata GenAI Powered Data Analytics Job Simulation
- Coursera — Google AI Professional Certificate
- Coursera — Google Data Analytics
- Cisco Networking Academy — Cisco Certified Network Associate
- Cisco Networking Academy — Apply AI: Analyze Customer Reviews
- Cisco Networking Academy — Data Analytics Essentials
- Cisco Networking Academy — Introduction to Modern AI
- Udemy — AI Cybersecurity Solutions: Overview of Applied AI Security
Open to Software Engineering internships (Backend & AI Systems) · Open source collaboration welcome


