Skip to content
View tonumayworkspace-creator's full-sized avatar
🎯
Focusing
🎯
Focusing
4 followers · 1 following

Block or report tonumayworkspace-creator

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.

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
tonumayworkspace-creator/README.md

👋 Hi, I'm Tonumay Bhattacharya

🚀 Generative AI Engineer | LLM • RAG • Agentic AI | FastAPI • Docker

⚡ Start Here (60-sec overview)

🧠 Profile Summary

+ 1+ year experience building production-grade LLM systems
+ Improved model accuracy by +14% and reduced false positives by 15%
+ Increased RAG answer relevance by 30–40%
+ Focused on real-world GenAI systems with deployment & evaluation

🧠 Core Expertise

+ Generative AI Systems (LLM • RAG • Agentic AI)
+ Production AI Pipelines (Design → Deploy → Evaluate)
+ Real-time AI APIs (FastAPI • Streaming • Multi-user)
+ Retrieval Optimization (Hybrid Search • Reranking • Chunking)

⚡ GenAI & LLM Stack

🛠️ Frameworks & Tools

🧩 Data, ML & Vector Systems

🔌 Backend & Deployment

🧪 Evaluation & Optimization

+ LLM Evaluation (Faithfulness • Grounding • Relevance)
+ Query Rewriting & Prompt Optimization
+ Hybrid Search (BM25 + Vector Retrieval)
+ Latency Optimization & Response Quality

🎥 Demo + 📊 Metrics + 🧠 Architecture

🎥 Demo

📊 Key Metrics

🧠 Architecture

User → Frontend → FastAPI → LangChain/LangGraph → Retriever → LLM → Response → Evaluation

🚀 Featured Projects

📊 Engineering Impact Dashboard

🧠 LLM & RAG Performance

  • 📈 RAG Answer Relevance: +30–40%
  • 🎯 Retrieval Precision: +20%
  • 🔍 Hybrid Search (BM25 + Vector)
  • 🧪 Evaluation: Faithfulness + Grounding

⚡ ML Performance

  • 📊 Accuracy: 68% → 82% (+14%)
  • 🚫 False Positives: ↓ 15%
  • 🧠 Feature Engineering Improvements
  • 📉 Metrics: F1 • AUC

🚀 System Performance

  • ⚡ 50+ concurrent users
  • ⏱️ Optimized latency
  • 🔌 FastAPI + Docker
  • 🔄 Real-time streaming

🤖 Agent Automation

  • 🔁 Manual effort ↓ ~60%
  • 🧠 Multi-step reasoning agents
  • 🔗 Tool integrations
  • ⚙️ Planner → Executor workflow

🧪 LLM Evaluation Approach

  • ✔ Faithfulness & Grounding (RAG validation)
  • ✔ Retrieval Recall & Precision
  • ✔ Latency monitoring (p50 / p95)
  • ✔ Hallucination reduction
  • ✔ Continuous evaluation (LangSmith)

🏗️ System Design Snapshot

User → Frontend → FastAPI → Orchestrator (LangChain / LangGraph)
     → Retriever (FAISS + BM25) → LLM (GPT-4o / Llama-3)
     → Response → Evaluation → Logs / Metrics

💼 Experience

  • AI Intern – SkillInfytech (2026–Present)
  • Data Science Intern – Codtech (+14% accuracy improvement)
  • ML Intern – iStudio (75–82% model accuracy)

🧩 Engineering Strengths

✔ Production-ready AI systems
✔ LLM + RAG + Agents
✔ API-first backend
✔ Evaluation-driven development

🏆 Achievements & Activity

💼 Experience

  • AI Intern – SkillInfytech (2026–Present)
  • Data Science Intern – Codtech (+14% accuracy improvement)
  • ML Intern – iStudio (75–82% model accuracy)

🧩 Engineering Strengths

✔ Production-ready AI systems
✔ LLM + RAG + Agents
✔ API-first backend
✔ Evaluation-driven development

📫 Connect

🐍 Contribution Graph

snake gif

Pinned Loading

  1. self-improving-rag self-improving-rag Public

    Self-Improving RAG System with Hybrid Search (BM25 + Vector), LLM Query Rewriting, RAGAS Evaluation, and Feedback Loop | Streamlit App

    Python 1

  2. credit-risk-decision-system credit-risk-decision-system Public

    Production-ready Credit Risk Decision Engine using ML, Expected Loss & FastAPI with Streamlit UI

    Python 1

  3. fraud-detection-real-time-system fraud-detection-real-time-system Public

    Real-time Fraud Detection System with Cost-Based Decision Engine (ML + FastAPI + Streamlit)

    Python 1

  4. rag-knowledge-assistant rag-knowledge-assistant Public

    Production-style RAG system with FastAPI, Streamlit, Ollama, and RAGAS

    Python 1

  5. RiskLens-Credit-Risk-Predictor RiskLens-Credit-Risk-Predictor Public

    End-to-end credit default risk prediction system with Explainable AI (SHAP), Flask web app, animated risk gauge, batch prediction, and admin dashboard UI.

    HTML 1

  6. FinGuard FinGuard Public

    FinGuard — credit-card fraud detection (Flask + scikit-learn)

    HTML 1