Headline

Applied AI Engineer | Senior Data Scientist | Top 10% Creator | Building Production RAG & Multi-Agent Systems | Local AI | (LangGraph, Ollama, AWS, Docker, Terraform)

About

Generative AI is not short on hype. It is short on engineers who care about whether a system actually works end-to-end. That is the problem I orient my career around. Senior Data Scientist → Applied AI Engineer Portfolio: https://sanjayg-portfolio.vercel.app/ AI Engineer's Field Guide & 1:1 sessions: https://topmate.io/sanjay_g24/ I specialize in Applied AI Engineering: multi-agent systems, RAG frameworks, and LLM pipelines that hold up in production, not just in demos. M.Tech from IIST (GATE, 99.5th percentile). My focus has shifted from generating outputs to engineering for performance: systems that reason, retrieve, verify, and act reliably under real-world constraints. What I've been building: → CiteMentor 2.0: an agentic RAG system with grounded responses featuring micro-royalty attribution framework for verifiable trust. Deployed end-to-end on AWS: Docker, Terraform, GitHub Actions CI/CD, ChromaDB, Langfuse for observability. → Local Invoice Intelligence: benchmarked a local Qwen3 14B extraction pipeline against GPT 5.5 on 297 DocILE invoices: 81.76% accuracy vs 83.04%, at zero API cost, and ahead of GPT on the hardest field (Gross Amount: 76.43% vs 70.71%). Stack: pdfplumber, Tesseract, Ollama, Pydantic. → SLM Hive: a local multi-agent, self-correcting Text-to-SQL swarm (Analyst, Architect, Auditor nodes) benchmarked on Spider: 77.4% accuracy vs. 53.4% for a single-model baseline. Core stack: Python, PyTorch, LangGraph/LangChain, Ollama, FastAPI, Vector DBs, AWS, Docker, Terraform, SQL. How I Think: I am drawn to problems at the intersection of complex logic and real-world constraints. I think in systems: how components interact, where failure propagates, and what makes a pipeline genuinely production-grade rather than just impressive. Outside the IDE: I am a lifelong student of systems and human psychology. Outside of engineering, you will find me diving into non-fiction, exploring personal finance, practicing Vedic wisdom, or lifting at the gym.

Location

Greater Bengaluru Area

AI stack

claude code

claude

codex

cursor

langchain

langgraph

Work experience

Senior Data Scientist at Gallagher, Mar 2024 - Present · 2 yrs 6 mos, Bengaluru, Karnataka, India, Leading GenAI and applied ML initiatives: 1. Generative AI Text Analytics: Built an LLM-based pipeline to automate sentiment and theme analysis across ~9k customer feedback entries spanning 25 branches, replacing manual review and enabling branch-level action plans. Estimated to cut manual review time by ~85% based on prior review throughput. 2. Predictive Talent Modeling: Built an attrition-risk model covering ~20k employees; validated against historical exits, the model captured 50–60% of actual departures within its top 10% of predictions.

Senior Data Science Analyst at Tiger Analytics, Jun 2022 - Mar 2024 · 1 yr 10 mos, Bengaluru, Karnataka, India, 1. Contributed to consulting engagements spanning marketing, industrial, and customer analytics for enterprise clients, working across GenAI-driven personalization, predictive maintenance, and large-scale segmentation. 2. Prototyped an XGBoost-based fault-forecasting model for a pharmaceutical client's capsule production line, conducting EDA on machine-fault data to surface leading indicators of downtime. 3. Explored LLM-assisted workflows for generating targeted marketing email drafts, prototyping GenAI-based personalization for customer segments. 4. Analyzed terabyte-scale customer data for RFM segmentation and persona profiling. Stack: XGBoost, LangChain, LLMs, PySpark, SQL, Azure Databricks

Data Scientist at Subex, Aug 2020 - May 2022 · 1 yr 10 mos, Bengaluru, Karnataka, India, My work involved: 1. Developing production-level codes using Python for building ML and DL models for applications in areas related to Scanned Document Parsing, Face Verification & Liveness Detection, Call Fraud Detection, Time Series Anomaly Detection. 2. Backend developer for the team working on "HyperSense AI Studio" - A No Code AI automation platform that is 36x faster and 20x more efficient than traditional coding methods. 3. Contributing novel solutions and strategies across projects and research works to solve challenging problems, hike productivity and enhance outcomes.

AI Intern at AlphaICs Corporation, May 2019 - Dec 2019 · 8 mos, Bangalore, Carried out a project to improve a SOTA DL-based Text-to-Speech system for my Master’s degree thesis and also published a paper on the same - “Natural Text-to-Speech Synthesis by Conditioning Spectrogram Predictions from Transformer Network on WaveGlow Vocoder”. Roles and Responsibilities: 1. Going through relevant papers to understand SOTA DL architectures for TTS. 2. Setting up GPU-supported docker containers for conducting model training on various open-source codes available. 3. Analyzing the results, coming up with new ideas to improve model performance, and implementing the code changes to train the modified model.

Education

Master of Technology - MTech, Digital Signal Processing, Indian Institute of Space Science and Technology

Bachelor of Engineering - BE, Electronics and Communications Engineering, Sjb institute of technology

Public links

github: https://github.com/sanjayg96

sanjayg-portfolio.vercel.app: https://sanjayg-portfolio.vercel.app/