Trayan Das

Senior AI/ML Analyst — Accenture

Sep 2024 – Present

• Privacy-Safe Analytics Agent Pipeline: Architected an end-to-end conversational analytics platform on Azure — incremental ingestion with PII/profanity redaction, a hybrid BM25 + vector AI Search index over 100k+ docs, and a multi-agent system — cutting retrieval latency ~75% and analyst time 60–70%.
• Enterprise Data Q&A Agent: Designed a text-to-SQL agent chain (Copilot Studio → AI Foundry) that translates natural-language queries into T-SQL executed directly against Lakehouse SQL endpoints, enabling analytics on 10M+ records in <30s.
• GenAI Business Doc Generator: Built an agentic, multi-agent RAG system with LangGraph and LoRA fine-tuned LLM for automated, context-aware business document generation on Azure, cutting drafting time by 60–70%.

Software Developer — Expro

Sep 2022 – Sep 2024

• Well Performance Analytics Platform: Built a PyTorch LSTM-autoencoder pipeline for anomaly detection on well-sensor time-series data, improving risk detection by ~20% and reducing analysis time by ~40%.
• Oil & Gas Insights Web Application: Built a PySpark KPI pipeline on Azure Databricks using PCA, K-means, and Random Forest to derive KPI features from equipment telemetry, reducing manual reporting by ~50%.

Data Scientist — Ganit Labs

Mar 2021 – Sep 2022

• Implemented predictive maintenance models on GCP for oilfield equipment achieving ~85% recall and reducing downtime by ~30%.

Sovereign-Policy-Assistant

Sovereign, offline RAG policy assistant with LangGraph, bilingual retrieval, Keycloak-gated fully offline confidentiality, and contradiction/staleness detection.

RTI-Engine

Multi-agent AI pay-transparency compliance system with LangGraph, tiered autonomy, MCP, and human-approval gating.

Autonomous Data Analyst

Multi-agent data analyst with LangGraph, evals, and observability.

Hybrid RAG from scratch

Hybrid RAG with from-scratch BM25, RRF fusion, FAISS, and RAGAS evals.

MARA — Multimodal Agentic Reasoning Assistant

Multi-agent AI system with LangGraph, RAG, vision, and analytics.

LLM from Scratch

Decoder-only transformer implemented in PyTorch with pretraining and fine-tuning.

Smart Compose System

Fine-tuned GPT-style model powering intelligent email composition.

LLM Summarization Pipeline

End-to-end summarization workflow using modern LLM techniques.

Company Brochure Generator

Automated content generation pipeline using large language models.

Languages: Python, SQL

AI/ML & GenAI: LLMs, RAG, Agentic AI, NLP, Machine Learning, Prompt Engineering

Frameworks: LangGraph, Flask, FastAPI, PyTorch

Cloud Services: Azure, GCP

Tools: Git, Azure DevOps, REST API

• Growth Catalyst Award — Accenture
• Promoted within 10 months at Ganit Labs
• Consistently delivered measurable AI/ML impact across enterprise projects

Claude Certified Architect – Foundations

AI-102: Microsoft Azure AI Engineer Associate

CS50AI: Introduction to Artificial Intelligence with Python — Harvard University

AZ-900: Microsoft Azure Fundamentals

AI-900: Microsoft Azure AI Fundamentals

M.Tech — University of Calcutta (2016–2018)

B.Tech — WBUT (2011–2015)