Dristanta Das Contact

Senior Data Scientist · Genpact, Bengaluru

I turn slow, manual data workflows into governed, self-service ones.

I design and ship Text-to-SQL, RAG and agentic systems with LangChain, LangGraph and PyTorch. Before Genpact, at UST, I built the provider search system (NER plus semantic search) that lifted search efficiency by 30%, still the result people ask about most.

Download resume (PDF) LinkedIn GitHub Email
Portrait of Dristanta Das

Production systems

Three production systems, with the measured result for each.

Conversational analytics: Text-to-SQL platform

Genpact

An agentic Text-to-SQL platform over a certified semantic layer. Business teams ask in plain English and get a governed, explained answer instead of waiting on a data-team ticket.

  • LangChain
  • LangGraph
  • Azure OpenAI
  • Agentic AI
  • Postgres
faster query formulation
65%
weekly queries from 50+ beta users
500+
agent task-completion rate
92%
cross-functional teams onboarded
12+

Invoice processing pipeline

Genpact

An LLM extraction pipeline for multi-format invoices across EMEA and NAM. Structured-output validation and a human-review queue for low-confidence fields replace blind automation.

  • LangChain
  • Azure OpenAI
  • Prompt engineering
  • Structured output
field-extraction accuracy
85%+
invoices processed
10,000+
less manual processing time
60%
languages across EMEA and NAM
8+

Provider search system

UST

An NLP and semantic-search system that maps layperson symptom descriptions to standardized medical terminology, so search resolves meaning rather than keywords. Built on NER and vector retrieval, before the GenAI wave.

  • NLP
  • NER
  • Semantic search
  • Open-source LLMs
boost in search efficiency
30%
faster data-preparation tasks
40%
gain in operational efficiency
25%

More projects

Personal and academic work across retrieval, audio, NLP, statistics and forecasting.

RAG QA bot

A Python QA bot with hybrid retrieval using BM25 and DPR. It integrates GPT-3.5-turbo, Phi-2 and Llama 3, with contextual compression to cut hallucinations.

40%higher accuracy than traditional search
60%fewer hallucinations
  • RAG
  • LangChain
  • BM25
  • DPR
  • LLMs

Resume analysis with spaCy

A resume scoring system that evaluates and ranks candidates against job requirements using spaCy.

  • NLP
  • Information extraction
  • spaCy

2.5D Visual Sound

Project report figure for 2.5D Visual Sound

Converts monaural audio to binaural by using the video, giving listeners a 3D sound sensation.

  • Audio processing
  • Computer vision
  • Deep learning

Bankruptcy prediction with LDA

Linear discriminant analysis plot for bankruptcy prediction

Linear Discriminant Analysis with dimensionality reduction to predict financial distress in companies.

  • LDA
  • Statistical modeling
  • Classification

FIFA 21 data analysis

FIFA 21 player statistics chart

Player attributes, market values and performance across positions and nationalities.

  • EDA
  • Sports analytics
  • Python
Zomato restaurant data analysis chart

Zomato food data analysis

Analysis of Zomato restaurant and delivery data covering customer preferences, pricing strategy and ratings.

  • Data visualization
  • R and Python
  • Business intelligence

Experience

Genpact

Bengaluru, Karnataka

Apr 2025 – Present

Senior Data Scientist

  • Architected and deployed the conversational analytics (Text-to-SQL) platform on LangChain and LangGraph, used by 50+ beta users across 10 concurrent sessions.
  • Engineered the agentic workflows behind it: state management, error handling and dynamic routing, scaled to 12+ cross-functional teams.
  • Led the automated invoice-processing pipeline on LangChain and Azure OpenAI.

See the results for both systems

UST

Kolkata, West Bengal

Jul 2022 – Mar 2025

Associate III Data Scientist Oct 2022 – Mar 2025

  • Developed a client-facing provider search system with NLP and open-source LLMs, boosting search efficiency by 30% and speeding data-preparation tasks by 40%.
  • Implemented NER and semantic (vector-based) search to map layperson language to medical terms, improving operational efficiency by 25%.

Associate II Data Scientist Jul 2022 – Sep 2022

  • Evaluated patient data in BigQuery, improving the client's understanding of adherence drivers by 45% across 10,000+ patients.
  • Designed a data anomaly-detection system for categorical healthcare data, in line with HIPAA and GDPR.

Skills

AI and ML
NLP, LLMs, NER, RAG, agentic AI, machine learning, deep learning, statistical modeling
Frameworks
LangChain, LangGraph, PyTorch, Hugging Face, scikit-learn, XGBoost, FastAPI, spaCy
Tools and cloud
Python, R, SQL, Git, Docker, BigQuery, vector databases, AWS, Azure (Azure OpenAI)
Design and governance
Architecture design, data-flow modeling, UML, anomaly detection, HIPAA and GDPR compliance