PavanSrinivas

AI / ML Engineer

Scroll

01 — Origin

Crafted in darkness.

The parts that make a model trustworthy are the parts nobody sees.

Retrieval that stays relevant. Evaluation that stays honest.

Pipelines that hold.

Origin00 / 03

02 — Macro details

Measured, not claimed.

  1. 01

    0.0%

    Semantic relevance

    LLM Systems

    Retrieval-Augmented Conversational AI

    A production-scale RAG system with validated semantic relevance and containerized deployment on AWS ECS.

    • RAG
    • Vector DB
    • AWS ECS
    • Docker
    • Python
  2. 02

    0.0%

    Intent accuracy

    Scalable NLP

    Consumer Complaint Classification

    An intent classifier for banking customer queries that routes most traffic automatically without sacrificing accuracy.

    • NLP
    • Text Classification
    • Python
    • Scikit-Learn
  3. 03

    0

    Model approaches compared

    Data-Driven Systems

    Credit Scoring & Customer Segmentation

    A credit risk scoring system built around a deliberate feature engineering strategy and tracked experimentation.

    • MLflow
    • Feature Engineering
    • Scikit-Learn
    • Pandas

03 — Specification

What it is made of.

LLMs & Generative AI
RAG Systems · LangGraph · Prompt Engineering · Fine-tuning · LLM Evaluation · Hugging Face · Vector Databases
ML & Data
PyTorch · TensorFlow · Keras · Scikit-Learn · Pandas · NumPy · Matplotlib
Languages & Tools
Python · SQL · FastAPI · Git · MLflow
Cloud & Infrastructure
AWS EC2 · AWS ECS · AWS Lambda · Docker · Containerization · Distributed Systems
Domains
Machine Learning · Natural Language Processing · Data Analytics

04 — Provenance

Where it was made.

Sep 2024 – Dec 2025Buffalo, NY

Data Engineer — Operations Lead

University at Buffalo

  • Designed and maintained a normalized database schema (people, inventory, transactions) with validation rules and data quality checks that ensured 100% transactional integrity.
  • Reduced query runtime by 40% by building real-time KPI dashboards with advanced SQL — window functions, CTEs, aggregations — and optimizing the underlying queries.
  • Engineered a production resource allocation pipeline using SQL window functions and time-series decomposition, validating business logic directly with stakeholders.
  • Improved asset utilization efficiency by 25% through ETL query refactoring and schema redesign.
  • Standardized and documented ETL pipelines for stock tracking, adding data quality gates and error-handling protocols that cut ingestion errors by 30% and shortened refresh cycles.

Education

  • M.S. in Engineering Science

    University at Buffalo, SUNY

    Buffalo, NY · Jan 2026

  • B.Tech in Artificial Intelligence

    SRM Institute of Science and Technology

    Chennai, India · May 2024

Certification

  • Claude with Amazon Bedrock

    Generative AI & LLM application development

    Jul 2026

  • AWS Cloud Practitioner

    Cloud platform deployment & AI services

    Jan 2026