Soumya Ranjan Maharana

Pro

Full Stack Developer

Junior Software Developer · Bankify Technology and Private LimitedLakshya Institute Of TechnologyBhubaneswar, OdishaAvailable from 2026-07-01 · full_time

Open to roles

GITMern Stack DevelopmentJavaJavaScript

GitHub LinkedIn

Chase the Cloud Challenge '26 (ISRO)

MeghDrishti AI — Diffusion-based INSAT cloud nowcasting for Chase the Cloud Challenge '26 (ISRO). LINKS • GitHub: https://github.com/Dynamicsoumya/Meghdrishtiai • API health: https://meghdrishtiai.onrender.com/health • Rubric map: docs/CHALLENGE_COVERAGE.md • Eval report: docs/evaluation_report.md • Dataset decisions: docs/dataset_card.md WHAT WE BUILT • 6 historical multi-channel frames (VIS/TIR/MIR/WV, 3 hours) → ≥2 future frames (t+30, t+60) • Conditional residual latent diffusion + ConvLSTM temporal encoder • Baselines: Persistence, Optical Flow, ConvLSTM • Metrics: MAE, RMSE, PSNR, SSIM, Cloud IoU/CSI, motion error • Dashboard: Time Machine, Reveal Actual, motion vectors, Storm Development Index, Futures wall, Judge Demo REGION SCOPE Bay of Bengal East Coast crop (lat 10–22N, lon 80–95E) — active convection + compute-aware; challenge allows region selection. Pipeline supports other MOSDAC crops. DATA NOTE MOSDAC INSAT-3D/3DR/3DS path documented (preprocessing/download.py). Demo bootstrap uses documented synthetic INSAT-like sequences so reviewers can run without MOSDAC credentials. LOCAL SETUP python -m venv .venv .venv\Scripts\activate # Windows pip install -r requirements.txt python scripts/generate_synthetic_data.py python inference/predict.py uvicorn dashboard.backend.main:app --port 8000 cd dashboard/frontend && npm i && npm run dev → http://localhost:3000 DASHBOARD SHORTCUTS D = Judge Demo | C = Cinema Mode | / = Commands | Reveal Actual = AI vs real wipe SCIENCE FRAMING Predicts satellite cloud evolution during INSAT gaps. Storm Development Index is decision-support only — NOT a lightning/IMD warning product. ONE SENTENCE The moment my model successfully predicted where the clouds would move next was when the Time Machine scrubber crossed NOW and the AI forecast continued the same convective cell northeast — then Reveal Actual confirmed the core stayed aligned.

5 media files · meghdrishtiai.vercel.appView

Brand & Creative Builder Award '26

ProofPath is a new brand for Brand & Creative Builder Award ’26 (New Brand Creation). WHAT TO REVIEW 1. Live product site (home) — brand experience, dual audience (talent + hiring teams), waitlist, projects, proof system 2. Case study —https://proof-path-jade.vercel.app/case-study 3. GitHub repo — https://github.com/Dynamicsoumya/ProofPath BRAND • Name: ProofPath (Proof + Path) • Tagline: From Skills to Work • Big idea: Proof > Promises • Campaign: No experience? Show your proof. STRATEGY Problem: Freshers have skills/certificates but face “experience required.” Insight: They don’t lack skills — they lack credible proof. Audience: 18–26 students & early-career professionals in India (persona: Rahul, 22) Promise: We don’t guarantee a job — we create verified proof that helps earn the first opportunity. APPLICATIONS (3+) 1. Website / landing 2. Mobile app screens (discover → progress → proof → opportunities) 3. Campaign + Proof Passport / physical touchpoints (in case study) SETUP No login or credentials needed. Open the live site and /case-study in a browser. AI DISCLOSURE AI was used for ideation/copy exploration. Final strategy, naming, identity, layouts, and creative execution were refined by the participant. Full disclosure is on the last page of the case study. Fonts: Space Grotesk + Inter (Google Fonts).

5 media files · proof-path-jade.vercel.appView

MethaneGuard AI '26

MethaneGuard AI — Every methane leak. Every second. Everywhere. End-to-end methane OS: detect → score → forecast → assign → repair → report. LIVE • App: https://methengaurdai-2.onrender.com • API: https://methengaurdai.onrender.com/api/health • Code: https://github.com/Dynamicsoumya/MethenGaurdAi • Docs: https://drive.google.com/file/d/1C76YoaaXLBdE-dMuNaAZ22O11q0ro9tB/view?usp=drive_link • Deck: https://docs.google.com/presentation/d/1Dk0JVUqty8J0IilQRk2n1oCzzEG09scW/edit?usp=drive_link&ouid=106482938334152467586&rtpof=true&sd=true Tip: Render free tier cold-starts ~30–60s — open API health first, then the app. JUDGE PATH (5 min) 1. Landing — 3D Earth with plumes, satellites, wind. 2. /login → Demo users → Admin / Analyst / Inspector (one click, no password). 3. /command — risk-ranked inspection queue, XAI, workflow, repair plan. 4. /detect — AI plume detection + Dice/IoU/mAP/F1 + TP/FP confusion. 5. /datasets — honest matrix: AVIRIS synthetic (primary) + real STARCOP fine-tune + TROPOMI .nc. 6. Bonus: /ceo · /twin · /inspector. DEMO USERS (/login → Demo users) • Admin — Priya Sharma (priya.admin@methaneguard.ai) → /admin • Analyst — Jordan Lee (jordan.analyst@methaneguard.ai) → /analyst • Inspector — Maya Chen (maya.field@methaneguard.ai) → /inspector DATASETS (honest) • Primary CV metrics: AVIRIS-like synthetic — Dice ~0.93 · IoU ~0.88 · mAP50 ~0.83 • Real AVIRIS-NG (STARCOP mini): fine-tuned — Dice ~0.83 · IoU ~0.70 · mAP50 ~0.70 (secondary) • TROPOMI / Sentinel-5P: sample NetCDF + ingest + UI — not used for U-Net train • False positives: classification head + val confusion (TP/FP/TN/FN) on /detect & /datasets WHY THIS PROJECT • Full ops product, not a notebook — mission-control UX. • Three real roles with separate workspaces. • Brief coverage: detect · noise rejection · severity · risk priority · dashboard · compliance impact. • Clear dataset honesty for judges. STACK Next.js · React · Three.js · Tailwind · Zustand · FastAPI · PyTorch MethaneUNet · scikit-learn Thank you — happy to walk through live.

5 media files · methengaurdai-2.onrender.comView

Product Design Builder Award '26

SafeNest AI is a secure family platform that organizes critical documents, medical information, and finances — and makes them instantly accessible during emergencies (e.g., a medical crisis where no one knows insurance details, medications, or emergency contacts). Research: Interviewed working professionals, parents, senior citizens, doctors, and insurance advisors. Key insight: trust in data security is the #1 adoption barrier, and renewal-date tracking is a recurring, universal pain point. Product thinking: Prioritized using Must/Should/Could — MVP focuses on Document Vault, AI Scanning, Expiry Reminders, and Emergency Mode, deliberately deferring hospital-record integrations that require institutional partnerships. Prototype: Clickable Figma covering onboarding, document scanning, dashboard, and the Emergency Mode hero flow. Usability tested with real task scenarios (find a policy, generate an emergency QR, check expiries). Background: Previously built at the Expiry Alert Hackathon — the same core expiry-tracking problem — now expanded into a full family safety platform, backed by MERN development experience for technical feasibility. AI Disclosure: AI tools assisted with brainstorming and content refinement. All research, prioritization, wireframes, interaction design, and testing were conducted independently.

5 media files · docs.google.comView

MSME Idea Hackathon 6.0 | Frontier Technology Innovation Challenge

Setu is a prototype for the MSME Idea Hackathon 6.0 (Frontier Technology / AI theme). It's an AI-powered web assistant that helps micro and small business owners discover eligible government schemes (CGTMSE, PMEGP, MUDRA, MSME Innovative Scheme) and auto-fills loan/subsidy application forms using document extraction (OCR + vision AI), reducing the paperwork burden that causes many eligible MSMEs to miss out on funding. The AI Assistant tab uses a live Claude API integration for real conversational Q&A on GST/compliance. The Auto-Fill tab uses AI vision to extract fields from uploaded documents (Udyam certificate, PAN card) and pre-populate application forms. Best viewed by opening the HTML file directly (AI features are functional when previewed in a Claude-connected environment; a walkthrough video is included above for full functionality).

4 media files · youtu.beView

ExpiryAlert '26 – AI-Powered Enterprise Document Expiry Management System

Organizations such as Deloitte, EY, KPMG, and Accenture manage hundreds of critical documents including vendor contracts, compliance certificates, insurance policies, and government licenses. Most organizations still track these records using spreadsheets, making it difficult to identify upcoming expirations. Missing an expiry can lead to compliance failures, financial penalties, operational disruptions, and audit risks. The challenge was to build a system that provides instant visibility into document status and proactively alerts teams before expirations occur. Process I first analyzed the problem statement and identified that the primary issue was not document storage but expiry visibility and proactive decision-making. I designed a dashboard-first experience focused on showing what needs immediate attention. Using the MERN stack, I implemented secure JWT authentication, CRUD operations for records, automatic status computation based on expiry dates, dashboard analytics, search and filtering, and an in-app notification system. I also added scheduled background jobs using node-cron to generate expiry alerts automatically. Throughout development, I prioritized simplicity, scalability, and a responsive user experience suitable for enterprise environments. Results The final solution provides a centralized platform for monitoring critical business records in real time. Users can instantly identify expired and expiring documents without manually reviewing spreadsheets. The dashboard significantly improves operational visibility, reduces manual tracking efforts, and enables proactive compliance management through automated alerts, search capabilities, analytics, and role-based record management. The architecture is scalable and production-ready for enterprise adoption. Reflection Given additional time, I would extend the platform with email and SMS notifications, role-based access control, audit logging, bulk Excel import/export, predictive analytics for renewal forecasting, and integrations with Slack, Google Calendar, and Microsoft Teams to further enhance enterprise collaboration and compliance workflows.

14 media files · expiry-alert-26-1.onrender.comView

MemoryVerse AI – Intelligent Personal Digital Identity Vault

Problem Students and professionals store certificates, projects, resumes, internship letters, and academic documents across folders, cloud drives, and platforms. Finding the right proof for job applications or interviews is slow and stressful. Existing storage tools only keep files — they do not understand achievements, skills, or how documents connect. Solution MemoryVerse AI is an AI-powered digital identity vault (not another Drive). It automatically categorizes documents, extracts skills and metadata, maps relationships between achievements, and supports semantic search in natural language so users can retrieve originals instantly. How it works I built a multi-stage AI pipeline that accepts documents, certificates, resumes, GitHub repositories, and portfolio links. The system extracts text (parsers + OCR where needed), generates embeddings with sentence-transformers, and stores vectors in ChromaDB. An intelligent categorization layer classifies content into Projects, Skills, Certifications, Internships, Achievements, and Academics. Extracted metadata builds a knowledge graph linking certifications → skills → projects → internships → career milestones, plus a journey timeline of growth over years. A semantic retrieval layer lets users ask questions like “Show my AI projects” instead of hunting by filename. The dashboard unifies the vault, timeline, relationship graph, and AI retrieval — with extras such as job-fit scoring, elevator pitch, and interview-ready proof packs. Outcome Successfully delivered an AI digital identity system that organizes achievements without manual folder sorting, enables natural-language retrieval with original file download, and visualizes career growth through an interactive knowledge map. Success metric: “I never have to search through folders again.” Next With more time: stronger multimodal document understanding, multi-user / campus access, richer date parsing, and a scalable cloud deployment with role-based access.

5 media files · memoryverseai-bnzygbwjtgmqkugtjj9zys.streamlit.appView

AMBULA - Healthcare Booking Platform

Millions of people in India struggle to find the right doctor and book appointments easily. Many clinics still manage appointments through paper registers, phone calls, or WhatsApp. This creates inefficiencies: doctors lack a clear digital view of their daily schedule, patients carry physical files between consultations, and double-bookings happen frequently. The challenge: Build a platform that enables patients to search doctors by specialization, book appointments in under 2 minutes, and allows doctors to manage schedules digitally while preventing overbooking through concurrent . Process Started with user research to understand pain points. Designed the system architecture using MERN stack for scalability. Built RESTful APIs with proper database constraints for concurrent booking handling. Implemented frontend with React for smooth UX on mobile. Key decisions: Used backend database locks to prevent double-booking (not just frontend validation). Created separate dashboards for doctors and patients. Added real-time slot availability updates. Used JWT authentication for security. What didn't work: Initial frontend-only booking validation failed with concurrent requests. Switched to backend-level transaction management which resolved the issue. Results Completed booking in <2 minutes on mobile device Prevented double-booking with concurrent request handling at backend Doctor dashboard showing all appointments + patient health records Patients can search 50+ doctors by specialization & location Consultation notes & prescriptions stored digitally Slot blocking feature for doctor holidays/leaves Mobile-friendly experience (responsive design) Secure login with JWT tokens Metrics: - Booking process: 90 seconds average - Concurrent request handling: Zero double-bookings under load - Mobile compatibility: 100% UI responsive Reflection Add real-time notifications via WebSocket when appointments are booked or cancelled. Implement automated appointment reminders via SMS/email to reduce no-shows. Add payment gateway integration for consultation fees. Implement doctor ratings & patient reviews system. Add prescription-to-pharmacy integration for direct medicine ordering.

9 media files · healthcarehub-frontend.onrender.comView

Student Expense Tracker — Python CLI for College Finance

College students in India lose track of monthly spending across UPI payments, canteen bills, mobile recharges, and travel — often overspending before month-end with no way to review where their money went. Existing apps require internet or account sign-ups. There was no simple, offline command-line tool a student could run instantly on any laptop. Process I started by mapping the 4 required features into a clean menu loop. I chose JSON file storage over in-memory lists so data survives restarts — a real student need. I built input validation for dates, amounts, and categories with helpful error messages. For the category chart, I used ASCII bar rendering to avoid any external libraries. I iterated on the budget warning to show a visual progress bar with percentage. I pre-loaded 5 realistic sample expenses (mess lunch, recharge, auto fare, canteen snacks, stationery) so reviewers see meaningful output immediately on first run. Results The program runs without errors, satisfies all 4 required features, and includes 2 bonus features (budget alerts and delete expense). It uses zero third-party libraries, stores data in a human-readable JSON file, and loads with 5 pre-seeded realistic expenses. Output is clearly formatted with Unicode box-drawing characters for readability in any terminal. Reflection With more time, I would add a monthly CSV export so students can share spending reports with parents or use them for scholarship applications. I would also add date-range filtering in the view screen and a recurring expense feature for fixed monthly costs like hostel fees.

1 media fileView

Queue Cure '26 — Live Clinic Queue Manager

76% of India's 1.5 million clinics run on paper token slips and shouting. Patients wait 2-3 hours with zero visibility into how long they'll actually wait. Doctors have no dashboard. Receptionists manage everything from memory — leading to errors, skipped tokens, and frustrated patients. There is no affordable digital solution built for small neighbourhood clinics. Process I started by breaking the problem into two user roles — receptionist and patient. For the receptionist, the key constraint was speed: adding a patient must take under 10 seconds. I used auto-token assignment and Enter key submission to achieve this. For the patient screen, the key constraint was trust: the wait time must come from real data, not a hardcoded guess. I implemented a rolling average that recalculates after every consultation using actual calledAt and servedAt timestamps. For real-time sync, I chose Socket.IO over polling because it gives true push updates with ~1ms latency — both screens subscribe to a single queue-updated event. I used MongoDB's atomic findOneAndUpdate with $inc to prevent race conditions when two receptionists add patients simultaneously. I also handled edge Results Delivered a fully working live clinic queue system with zero page refresh. Receptionist can add a patient and assign a token in under 5 seconds. Patient screen updates within milliseconds of calling next token. Wait time is computed from real consultation data using a rolling average — not hardcoded. The system handles concurrency, server restarts, and edge cases gracefully. Built and deployed within the hackathon timeline using MERN stack + Socket.IO. Reflection Add separate role-based dashboards for receptionists and patients with personalized features.

2 media filesView

BusAlert Pro — Real-Time Bus Tracking with AI Delay Prediction

Public bus systems often lack real-time predictability, making it difficult for users to plan their travel efficiently. Passengers frequently face uncertainty in bus arrival times, missed buses, and inconvenient delays in reordering or rebooking their travel plans. This leads to poor user experience, wasted time, and reduced trust in public transport systems. Process 1. User Login / Entry User opens the app Logs in or continues as guest System detects current location 2. Search & Select Bus Route User enters destination System shows available buses + routes Displays: Bus number Estimated arrival time (ETA) Live tracking status 3. AI Prediction Layer System analyzes: Live GPS data Traffic conditions Historical bus delay patterns AI predicts: Accurate arrival time Delay probability 4. Travel Decision Screen User sees: “Bus arriving in 8 min (delayed by 3 min)” Alternative options: Next bus Faster route Nearby stop suggestion 5. Frictionless Reorder Flow (Core Feature) If user misses or wants to change plan: No need to re-enter details 6. Confirmation System Results 1. Improved Travel Experience Users get real-time and accurate bus arrival updates Reduced uncertainty about bus timing Less stress while planning daily commute 2. Faster Decision Making One-tap reorder / reschedule flow reduces manual effort Users can quickly switch buses or routes without restarting search 3. Reduced Waiting Time AI predicts delays and suggests alternatives early Helps users avoid missed buses and long waiting at stops 4. Smarter Public Transport Usage Better distribution of passengers across buses Reduced overcrowding in peak routes Optimized utilization of available buses Reflection Instead of normal rebooking: If bus is late or missed → system automatically rebuilds full travel plan Includes: Next best bus Walking route Nearby stop change

2 media filesView

This is Soumya’s work on Wooble.

Build a profile that shows what you can do — and share it anywhere.

Build yours