Skip to content
View prathamkashyap's full-sized avatar
🍀
Chasing midnight thoughts
🍀
Chasing midnight thoughts

Highlights

  • Pro

Block or report prathamkashyap

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
prathamkashyap/README.md
Crimson Archive — overview, quest log, loadout — online at VIT Bhopal The Crimson Moon Warrior — cyber samurai standing before a blood moon, katana drawn

Class: The Systems Forger  ·  Guild: VIT Bhopal — AI & ML  ·  8 Quests Cleared  ·  Status: Open for New Quests

Portfolio  ·  GitHub  ·  LinkedIn  ·  Email

The Crimson Moon Warrior awakening under a blood moon

Character Record — Welcome to the archive

Character record — AI Systems Engineer profile, Crimson Moon Warrior

Current campaign — what is being forged

  • AI Systems LabMAIN QUEST · S-RANK
    Inference gateway with SingleFlight deduplication, circuit breakers, and hybrid BM25 + dense retrieval fused with Reciprocal Rank Fusion.
  • LibrisSIDE QUEST · A-RANK
    Enterprise-style library backend on Spring Boot: layered architecture, Google OAuth2/OIDC SSO, dual-theme frontend, CI-enforced coverage.
  • AthenaeumAISIDE QUEST · B-RANK
    RAG-grounded study platform that turns PDFs into quizzes and flashcards, with SM-2 spaced repetition driving the review queue.
  • Iron pathSYSTEMS DESIGN
    REST contracts, caching, indexing, and write-heavy consistency.
  • Sight & voiceFIELD OPERATIONS
    YOLOv8/11, SAM, OpenCV, Whisper — realtime inference with graceful degradation.

Loadout equipped

🟨 LEGENDARY — AI Systems
Python FastAPI Redis-style caching RRF retrieval Prometheus

🟪 EPIC — Backend Forge
Java 21 Spring Boot Node.js Express Flask

🟦 RARE — Vision Core
PyTorch TensorFlow OpenCV YOLOv8/11 SAM Whisper

🟦 RARE — Data Sense
PostgreSQL MySQL MongoDB Firebase Docker Linux Git

LanguagesC++ Python Java JavaScript TypeScript SQL
WebReact Vite FastAPI Spring Boot Node.js
InfrastructureDocker Git Linux GitHub Actions

Quest log — eight quests

MAIN QUEST · LEGENDARY

An inference gateway forged to survive the live server, not just the tutorial zone.

An engineering reference for AI infrastructure, built as a gateway rather than a notebook. TTL+LRU caching, SingleFlight deduplication, micro-batching, circuit breaking, and failover across Anthropic, OpenAI-compatible backends (Ollama, vLLM, DeepSeek, Groq), plus a deterministic MockProvider. Hybrid BM25 + dense-vector retrieval with Reciprocal Rank Fusion, a RAG evaluation suite, an interactive playground, Prometheus metrics, and 40 tests.

Benchmarked: 200 concurrent duplicate prompts drop from 297.7ms to 45.5ms wall time via SingleFlight — a 99.5% deduplication rate.

Python FastAPI Caching RRF Prometheus · Repository

  Libris

SIDE QUEST · EPIC

A library guild's records vault, armored with the same layers a real fortress needs.

Enterprise-style library backend, built the way production systems are built. Spring Boot 3.5 layered REST (Controller → Service → Repository), MySQL, Flyway, Ember Dark / Verdigris Light frontend. BCrypt sessions plus opt-in Google OAuth2/OIDC, Admin / Librarian / Student RBAC, CSRF protection, JSON logging, Swagger/OpenAPI. 20 tests across 4 classes, JaCoCo ≥70%.

Java 21 Spring Boot MySQL Flyway Docker · Repository · Live

SIDE QUEST · EPIC

Turns a graveyard of forgotten PDFs into a study companion that remembers what you keep forgetting.

RAG-grounded study system: chunked PDF ingestion, semantic retrieval, Groq Llama 3.3 70B generation, SM-2 spaced repetition, exponential-decay retention. Redis/BullMQ jobs with typed enqueue errors and retry-safe validation. 118 unit tests, 21 integration tests, Playwright E2E, GitHub Actions CI.

React Node.js MongoDB Redis Groq · Repository

FIELD QUEST · RARE

One watchtower over the whole shop floor — faces, shelves, reviews, and a chatbot that actually answers back.

End-to-end retail intelligence behind one gateway: OpenCV face encodings (never raw images), MobileNetV2 product classification, TF-IDF + logistic regression review sentiment, hybrid rule + ML-intent chatbot. React + Vite dashboard, webcam capture, Chart.js analytics.

OpenCV React · Repository · Live

  CiviQ

FIELD QUEST · RARE

Reads a city's complaints like a battle map, so the right crew gets sent to the right street.

Geospatial ML for civic-complaint triage. KNN spatial-density features, DBSCAN hotspots, XGBoost classifier, 5-fold CV on 1,600 synthetic + 5,469 NYC 311 records — 95.7% accuracy. Spatial-density features alone lifted Decision Tree accuracy by 12.9 points. React/Firebase reporting with Leaflet hotspot maps.

React Firebase XGBoost · Repository

FIELD QUEST

An assistant that stays asleep until it sees a face — then listens.

Face-gated voice assistant that triggers only on visual presence. OpenCV Haar-cascade gating, Whisper transcription, deterministic 5-category intent router with fallbacks.

Python OpenCV Whisper · Repository

FIELD QUEST

A shopping list summoned by voice alone, in four languages, with no server standing behind it.

Browser-native voice shopping list — no backend, no login. Web Speech API parser for add / remove / quantity / brand / price, with history and seasonal suggestions. English, Hindi, Telugu, and Tamil, plus text fallback.

JavaScript Web Speech API · Repository

FIELD QUEST

Reads a leaf like a diagnosis chart — one model finds it, the other names what's wrong.

SAM isolates leaf and stem; custom YOLOv11 detectors diagnose wheat, tomato, and cotton disease.

YOLOv11 SAM TensorFlow Lite · Repository

Research Codex — notes from the field

  • GA-ET-IVCFS-ANFIS — Neuro-fuzzy product ranking under high uncertainty, co-authored with faculty (R² = 0.98).
  • Civic complaint prioritization — Hybrid synthetic + NYC 311 evaluation for geospatial civic triage.
  • Multimodal AI Assistant — Solo-authored IEEE-style research on realtime vision–speech co-activation.
  • Parkinson's biomarker mining — Acoustic feature pipeline with optimized SVM (F1: 0.935, ROC-AUC: 0.945).

Trophy Room — milestones unlocked

8+ quests shipped, 4 research tracks, 400+ problems solved, CodeVita global rank 3633

Checked against live profiles rather than an outdated recap.

⚔️ LeetCode400+ problems cleared (402 in C++ alone)
⚔️ Codeforces — Rank 815 (Newbie) · 288 cleared all-time
⚔️ CodeChef — 922 rating, 1-star Div 4 · DSA 1067 · Diamond League · 335 cleared
⚔️ HackerEarth — Top 2% in Basic Programming
⚔️ HackerRank — Certified Developer

Full trial history: competitive-programming.

Seals

  • TCS CodeVita Season 13 — Global rank 3,633
  • Oracle Certified Foundations Associate — Agentic AI, Oracle University
  • NPTEL (IIT Kharagpur) — Cloud Computing (Elite + Silver, top 2%) and Introduction to IoT (Elite + Gold, top 5%)

Field notes — active quest

  •   Currently reading: Designing Data-Intensive Applications by Martin Kleppmann. Every page is a blueprint.
  •   Current rabbit hole: Model Context Protocol (MCP) and autonomous coding-agent architecture. The next frontier.
  •   Engineering bias: End-to-end usable systems over isolated model demos.

Combat log — GitHub telemetry

GitHub stats GitHub streak

Guild connections — send a signal

Crimson Archive — cleared in the field, not the tutorial — open for new quests

Pinned Loading

  1. ai-systems-lab ai-systems-lab Public

    Experiments and prototypes in AI systems engineering, backend infrastructure, model serving, and applied machine learning.

    Python

  2. AthenaeumAI AthenaeumAI Public

    AI-powered quiz and study platform that generates MCQs, flashcards, and PYQ-style questions directly from uploaded study material using a retrieval-based approach.

    TypeScript

  3. libris libris Public

    A full-stack Library Management System built with Spring Boot, Spring Security, MySQL, and Vanilla JavaScript.

    Java 1

  4. civic-lens civic-lens Public

    A civic intelligence platform for reporting, prioritizing, and visualizing urban issues

    Jupyter Notebook

  5. evidence-driven-dataset-intelligence evidence-driven-dataset-intelligence Public

    Evidence-aware dataset intelligence and recommendation system for finding, evaluating, ranking, and diversifying public datasets for machine learning tasks.

    Python

  6. competitive-programming competitive-programming Public

    Multi-platform competitive programming solutions, reusable DSA templates, and algorithmic revision notes in C++, Python, and Java.

    C++