Vetted Retrieval-Augmented Generation Professionals

Peeyush Dyavarashetty

Screened ReferencesModerate rec.

Intern AI/ML Engineer specializing in GenAI, LLMs, and agentic RAG systems

Miami, FL2y exp

“AI/LLM practitioner who built a GPT-2-like language model from scratch at the University of Maryland using PyTorch and multi-GPU distributed training, with experiment tracking in Weights & Biases. As an AI Operations intern at ScaleUp360, delivered multiple production-style AI agent automations (Gmail classification and Fireflies-to-Claude workflows that extract and assign CEO tasks) and set up measurable evaluation using test cases and classification metrics.”

Pavan Devulapalle

Screened ReferencesModerate rec.

Mid-level Software Engineer specializing in cloud platforms and AI-integrated full-stack development

Seattle, WA3y exp

“Backend engineer who built Flask-based internal APIs supporting GenAI-driven provisioning/diagnostics (Outpost/AWS Outposts-like environment), with deep hands-on optimization across Postgres/SQLAlchemy (2s to <200ms endpoint improvement). Experienced integrating ML/LLM workflows via AWS SageMaker and Bedrock, and designing multi-tenant isolation plus high-throughput Redis-backed background task pipelines (minutes to seconds).”

“Platform/infra engineer with strong ownership of Kubernetes on VMware and day-to-day hybrid on-prem-to-AWS operations. Has hands-on experience automating infrastructure delivery with Terraform/Ansible/CI-CD, and has resolved real production issues spanning CSI storage reattachment during upgrades, vSphere storage-latency performance degradation, and hybrid connectivity/routing failures with improved validation, monitoring, and failover.”

Gagan Mundada

Screened

Intern Machine Learning Engineer specializing in multimodal AI and evaluation benchmarks

San Diego, CA2y exp

“ML-focused candidate with beginner ROS/ROS2 experience (custom pub-sub nodes; TurtleBot3 SLAM simulation debugging via topic inspection and transform/orientation checks). Has research/project exposure to LLM training approaches (GRPO with pseudo-labels using Hugging Face TRL on Qwen/Llama) and uses Docker/Kubernetes + CI/CD to run ViT saliency-attention/compression workloads on UCSD Nautilus infrastructure.”

Vignesh Shanmugasundaram

Screened

Junior Software Engineer specializing in full-stack development and applied ML

New York, NY2y exp

“Full-stack engineer with experience at Zoho and Amazon who has owned production systems end-to-end, including a monolith-to-microservices migration using Kafka and Cassandra that improved search latency ~25% and increased throughput without data loss. Also built a hackathon project (Buildwise) into a sold product for a construction company (AI-driven document compliance checks) and shipped an IoT-based parking availability MVP in 3 weeks.”

Poorna Pedapudi

Screened

Mid-Level Software Engineer specializing in distributed backend systems and cloud-native microservices

Seattle, WA5y exp

“Software engineer focused on data platforms and applied LLM systems: built an internal data quality monitoring layer to catch silent data drift and iterated post-launch after finding ~30% false-positive alerts, reducing noise via dynamic baselines and improved structured logging. Also shipped a production RAG-based internal knowledge assistant over Jira/Confluence with citations, confidence-based fallbacks, and nightly automated evals to prevent regressions.”

Aishwarya Sheelvant

Screened

Junior Backend & Data Engineer specializing in cloud infrastructure and ML pipelines

Atlanta, GA2y exp

“Built a GenAI/RAG-based ESG questionnaire-answering agent at C3.ai, including a React dashboard with role-based access and human-in-the-loop verification by showing supporting source paragraphs. Reported outcomes included cutting a 4–5 week manual process down to about a week (~90% labor reduction) and a client-reported ESG rank improvement from 7th to 3rd.”

“Built and productionized a multi-agent LLM analytics assistant at eBay that routes natural-language questions to retrieval or text-to-SQL, dynamically retrieves relevant schemas via a vector DB, and executes against a data warehouse. Drove a major quality lift (text-to-SQL accuracy 60%→85%) and materially reduced time engineers/PMs spent getting data insights through strong eval/monitoring, tracing, and reliability-focused design (schema retrieval, strict JSON outputs, retries/clarifications).”

“Built a production GenAI support agent at Amazon for FBA on-call operations, using Bedrock, Lambda, RAG, and confidence-based human fallback to safely automate ticket triage. The system materially reduced ticket volume and manual workload while improving MTTR, showing strong depth in reliable LLM agent architecture under real operational constraints.”

Tara Munjuluri

Screened

Junior Software Engineer specializing in full-stack and AI systems

Ames, IA3y exp

“Backend-focused engineer with end-to-end ownership experience on internal platforms at John Deere, including a workforce and skills system that cut manual review time by 40%. Brings a strong reliability and compliance mindset across Java/Python microservices, AWS infrastructure, and production operations, and has also built an LLM-powered RAG system over 1M+ records with emphasis on grounded outputs and observability.”

“Backend/full-stack engineer (5+ years) with Shopify experience integrating LLM/RAG workflows into production APIs. Owned a Python TensorFlow Serving inference pipeline connected to Java microservices via gRPC, optimizing tail latency at ~10k concurrent load and improving retrieval relevance with embedding and evaluation work. Strong Kubernetes/EKS + GitOps/CI/CD background, including monolith-to-microservices migrations and event-driven streaming patterns.”

Benjamin Kozel

Screened

Mid-level Machine Learning & Software Engineer specializing in RAG systems and ML infrastructure

Atlanta, GA4y exp

“Built and deployed an in-house RAG LLM system ("MONTY") using LLaMA 3B + FAISS to help teams quickly understand long internal/external specifications. Delivered usable production performance despite severe compute limits (single RTX 3080) by tuning retrieval/reranking and model choice, and is planning a LightRAG/knowledge-graph rewrite to improve accuracy and latency.”