class AdarshSaurabh:
def __init__(self):
self.education = "M.Tech in Signal & Image Processing @ NIT Rourkela"
self.undergrad = "B.Tech in CSE @ Guru Ghasidas University (8.5 CGPA)"
self.qualifications = ["GATE CSE 2025", "GATE DA 2025"]
self.specializations = [
"High-Throughput ML Pipelines & GBDT GPU Acceleration",
"Local LLM Inference Serving & Multi-Agent Systems",
"Computer Vision & Edge Device Deployment"
]- 🎓 Research & Degree: M.Tech candidate (Signal & Image Processing), National Institute of Technology, Rourkela (2025–2027).
- 🏆 Competitive ML: Built 24.2M-record Entity Resolution pipeline for Amazon ML Challenge 2026 (
Macro F0.5 = 0.9719). - 🤖 GenAI Systems: Built adversarial multi-agent document intelligence workflow using LangGraph + Gemini 2.5 Pro.
- 🥇 Hackathons: Rank 1 @ Girlscript "Hack4CG" (Facial Emotion Recognition); Intermediate Winner @ Awaiken Hacks (MediaPipe CV).
| Layer | Technologies |
|---|---|
| Core Languages | Python, C++, C, SQL, TypeScript |
| ML & Data Systems | Polars, CatBoost (GPU), LightGBM, RapidFuzz, Scikit-Learn, PyTorch, OpenCV |
| LLMs & GenAI | LangGraph, LiteLLM, llama.cpp, Gemini API, Hugging Face, Streamlit |
| Backend & Cloud | FastAPI, Django REST, Next.js, Supabase, PostgreSQL, Docker |
| Project / Repository | Focus & Architecture | Stack | Links & Key Results |
|---|---|---|---|
| Amazon ML Challenge 2026 | 24.2M-record Entity Resolution engine with multi-view blocking | CatBoost GPU, LightGBM, Polars, RapidFuzz | Code • Macro F0.5 = 0.9719 |
| IBYD Warehouse PathMapper | Heuristic routing system for dynamic warehouse mapping (10,000×10,000 units) | Python, Graph Algorithms, Heuristics | Video Demo • <0.5s for 10,000+ points |
| Uplan AI | Adversarial multi-agent document verification with Gemini 2.5 Pro | LangGraph, Gemini API, Streamlit, Python | Hugging Face Space • 85% manual audit cut |
| local-gpu-llm-hub | Stateless local GPU LLM inference server + LiteLLM proxy | FastAPI, llama.cpp, CUDA, Web UI | Code • OpenAI/Claude API drop-in |
| Alternative Data Radar | Automated pre-earnings web signals & corporate health index | TypeScript, Next.js, Supabase, Bright Data | Code • Real-time Recharts dashboard |
| Apna Gold SaaS | Multi-tenant B2B2C gold platform with KYC verification & tenant isolation | Django REST API, React Vite, JWT | Live Platform • Production SaaS |
| CodeAtlas | Zero-dependency AST architecture graph & performance bottleneck analyzer | Python ast, Graph Algorithms, Mermaid.js |
Code • AST structural analysis |
| K-HUKI | Keyframe extraction engine with HOG features & unsupervised clustering | Python, Computer Vision, OpenCV, ML | Project Report • 96.45% acc, 11x faster ResNet |
| Wildlife-Tracking | Edge-network real-time hybrid animal tracking | PyTorch, Computer Vision, Edge AI | Code • Real-time edge inference |
Click to expand / collapse resume details
- National Institute of Technology, Rourkela (2025 -- 2027)
Master of Technology in Signal and Image Processing --- CGPA: 8.28 - Guru Ghasidas University, Bilaspur (2020 -- 2024)
Bachelor of Technology in Computer Science & Engineering --- CGPA: 8.50
- Machine Learning Intern --- Autobot Robotics (May 2023 -- Aug 2023)
- Mentored junior interns in core ML, NumPy, Pandas, and exploratory data analysis.
- Built and deployed 4 production-grade ML/DL model microservices in Django for robot integration.
- Team Leader (Academic Alliance Program) --- Ziroh Labs (Jan 2022 -- Feb 2022)
- Led a 5-engineer team to deliver core image processing pipelines ahead of schedule.
- Languages:
Python,C,C++,SQL,TypeScript,HTML/CSS - ML & Systems:
Polars,CatBoost (GPU),LightGBM,RapidFuzz,PyTorch,Scikit-learn,OpenCV,CUDA - LLMs & GenAI:
LangGraph,Gemini 2.5 Pro / Flash,LiteLLM,llama.cpp,FastAPI,Streamlit,Supabase - Web & Data Platforms:
Django REST API,Next.js,React Vite,JWT,PostgreSQL,Docker
- Amazon ML Challenge 2026 --- Entity Resolution Engine (2026) --- CatBoost GPU, LightGBM, Polars, RapidFuzz --- Code •
Macro F0.5 = 0.9719 - Uplan --- Adversarial Document Intelligence Pipeline (2025) --- LangGraph, Gemini 2.5 Pro, Streamlit --- Hugging Face Space • 85% manual audit reduction
- Alternative Data Radar (2026) --- Next.js, TypeScript, Supabase, Bright Data --- Code • Corporate health index
- IBYD Technology --- Warehouse PathMapper (2023) --- Python, Heuristic Routing --- Video Demo • Sub-0.5s routing on 10,000×10,000 grids
- Apna Gold Solutions --- Multi-Tenant SaaS Platform (2025) --- Django REST, React Vite, JWT --- Live Platform • Production B2B2C
- K-HOG Unsupervised Keyframe Identifier (K-HUKI) (2023) --- Python, OpenCV, HOG --- Report • 96.45% accuracy, 11x faster ResNet
- local-gpu-llm-hub (2026) --- FastAPI, llama.cpp, CUDA --- Code • Local GPU inference server
- CodeAtlas (2026) --- Python
ast, Mermaid.js --- Code • AST architecture analysis - Wildlife-Tracking (2025) --- PyTorch, Edge AI --- Code • Real-time animal tracking
-6C63FF?style=for-the-badge&logo=adobeacrobatreader&logoColor=white){kind=icon}
