Looking for a job

Pooja Singari

@poojasingari

Message

AI/ML Engineer at TEKsystems (HPE R&D Client), building LLM-powered systems that cut clustering latency by 75%.

India

Message

At TEKsystems (HPE R&D Client), I engineered an LLM-powered ICR pipeline for 2.3M+ cases, cutting clustering latency by 75% and API latency by 65%. I also productionized a Case Evaluation engine for SFDC AI Assist that ranks case-engineer pairs across 10 dimensions.

Previously, I developed LLM features for semantic search and Agentic RAG at StartUs Insights, and optimized quantized models for edge devices at IoTIoT. My projects include a lip-reading model that achieved 96.5% accuracy and a LangGraph-based agent with human-in-the-loop approval gates.

Experience

Work history, roles, and key accomplishments

Current

AI/ML Engineer

Current

TEKsystems

Oct 2024 - Present (2 years 1 month)

Engineered LLM-powered ICR pipeline for 2.3M+ cases, cutting clustering latency by 75% and API latency by 65%. Designed and productionized Case Evaluation engine for SFDC AI Assist and built pre-order Quote-to-Cash validation gate using Azure Document Intelligence and Neo4j.

LLM ICR DuckDB Azure Document Intelligence Neo4j fastAPI Kubernetes Langfuse Oracle PostgreSQL SQL Server RAG Text To SQL

AI/ML Engineer

StartUs Insights

Jul 2024 - Sep 2024 (2 months)

Launched a new semantic search feature by developing and optimizing an LLM pipeline to generate high-quality user-facing feature tags. Architected an Agentic RAG workflow with an intelligent query router, improving search accuracy and retrieval precision.

LLM Semantic Search RAG Agentic Workflows Query Routing Python

AI Intern

IoTIoT

Feb 2023 - May 2023 (3 months)

Optimized and deployed quantized Hugging Face models and a C++ OpenCV API on edge devices, achieving 25% lower inference latency and enabling real-time computer vision under memory constraints.

Hugging Face OpenCV C++Edge AI Model Quantization Computer Vision