Available for Internships

Debasmita Chatterjee

AI Engineer · LLM Systems · Applied ML

B.Tech undergraduate building production-grade AI systems — from LLM calibration frameworks to governance platforms and RAG-based automation pipelines.

0 LLM responses evaluated
0 % ECE reduction achieved
0 Live deployed projects
Scroll to explore

Where I've
Worked

Outlier.ai Aug 2025 – Jan 2026
AI Trainer – LLM Evaluation (RLHF) · Remote
  • Evaluated 1,500+ LLM-generated responses using rubric-based scoring frameworks — identifying hallucinations, logical inconsistencies, and failure modes to support model reliability and RLHF training pipelines.
  • Contributed to LLM performance assessment across reasoning, multilingual, and knowledge tasks by maintaining structured annotation taxonomies and quality benchmarks — directly informing model optimization decisions.
  • Performed fine-grained comparative analysis and preference ranking of model outputs, providing actionable feedback to engineering teams to improve LLM accuracy, clarity, and response quality.

What I've
Built

01

LLM Confidence Calibration & Overconfidence Analysis

Production-grade statistical framework for diagnosing and correcting overconfidence in Mistral-7B and Phi-2 via logit-level extraction and post-hoc temperature scaling.

~62% ECE Reduction 18.4% → 13.6% Hallucination 500 BoolQ Samples
Python PyTorch HuggingFace SciPy
02

AI Agency Workflow Automation Platform

End-to-end AI automation platform with gradient-boosted ML scoring engine, FAISS-powered RAG pipeline, and LLM-based proposal generation via REST API.

6 Automated Stages Real-time Inference Live Deployed
Python FastAPI Scikit-learn FAISS Streamlit
03

Enterprise AI Governance & Risk Intelligence Platform

Scalable AI governance platform evaluating vendors across 6 risk dimensions using weighted scoring, Monte Carlo simulation, and role-based audit workflows.

300 Monte Carlo Iterations 3 User Roles Docker · GCP Cloud Run
Python Streamlit SQLite Docker GCP Cloud Run ↗

Verified
Credentials

Technical
Arsenal

Languages & Frameworks
Python PyTorch HuggingFace Transformers Scikit-learn FastAPI Streamlit
AI / ML
LLM Evaluation RLHF RAG FAISS Temperature Scaling Calibration Prompt Engineering
Data Engineering
SQL SQLite Pandas NumPy SciPy Data Pipelines
Cloud & Infrastructure
Oracle Cloud Docker GitHub Actions GCP Cloud Run ↗ soon Git Jupyter

Academic
Background

Lovely Professional University
B.Tech in Computer Science & Engineering
2022 – Present
Punjab, India
Let's Work
Together

Open to AI internships, research collaborations, and interesting problems.