Skip to content

Luv Patel

Forward Deployed Engineer • l***********@gmail.com • 469****612 • drivetube.ai/•••••

Professional Summary

Forward Deployed Engineer with 0 years of experience deploying and operating LLM-driven products and retrieval systems for production use. Practical experience building LangGraph agent platforms, hybrid dense + sparse retrieval with Qdrant, evaluation pipelines on Kubernetes, and customer-facing observability and grounding layers that reduce latency and catch regressions pre-release.

Technical Skills

Programming Language: Python,Go,TypeScript,JavaScript,SQL
Frontend Technologies: React,Next.js,React Native
Backend Technologies: FastAPI,NestJS,Node.js,OIDC
Databases: PostgreSQL,MongoDB,Redis,Ray
Cloud Platforms: AWS,S3,SQS,IAM,Google Cloud Platform,SSO,CLI tooling,Distributed tracing,Structured logging,Automated alerting
DevOps & Infrastructure: Terraform,Kubernetes,Docker,GitHub Actions,Continuous Integration,Continuous Deployment
APIs & Integration: REST APIs
Machine Learning & AI: Ollama,Hybrid dense-sparse retrieval
Generative AI & LLMs: LangGraph
Vector Databases & RAG: Qdrant,Vector Search,Reranking,Retrieval-Augmented Generation,RAG,MLflow

Work Experience

SPREE (Kingship AI)
Remote
Founding AI Software Engineer
Sep 2025 – Present
Worked on a production LLM agent platform and associated backend services for a consumer-facing AI product; focused on orchestration, latency, and observability.
Tech Stack: LangGraph, FastAPI, PostgreSQL, Kubernetes, AWS SQS, SSO, OIDC, Distributed tracing, Structured logging
  • Architected and shipped a production LLM agent platform from zero using LangGraph orchestration, FastAPI microservices, and PostgreSQL schemas, enabling multi-step agent workflows for end users.
  • Re-architected synchronous request flows into async event-driven pipelines using AWS SQS and background workers, reducing API latency by 25% and increasing user engagement by 15%.
  • Designed and deployed platform on Kubernetes with SSO/OIDC integration for secure customer access and automated deployment pipelines, shortening deployment time for releases.
  • Built an evaluation and observability layer with distributed tracing and structured logging to detect agent regressions pre-release, reducing post-release incidents and cutting onboarding time for engineers.
  • Implemented grounding and verification checks across agent outputs to attach source citations, improving answer traceability and lowering user-reported hallucinations.
  • Led performance tuning and capacity planning for production services, establishing SLAs and automated alerting that improved platform stability during peak usage.
Faradyne
Co-Founder and Technical Lead
Jan 2026 – Present
Led development of an AI product converting unstructured technical electronics documents into structured, source-cited outputs; built retrieval and evaluation infrastructure.
Tech Stack: LangGraph, Qdrant, Cross-encoder reranking, Kubernetes, Embeddings and vector search, Python
  • Owned the full AI stack for a live product: built LangGraph pipelines to ingest schematics, KiCad/Gerber files, and PDFs into structured, source-cited outputs for engineering workflows.
  • Engineered hybrid dense + sparse retrieval using Qdrant for embeddings and added cross-encoder reranking to improve precision of retrieved evidence used for model grounding.
  • Implemented a grounding-verification stage that anchors every model claim to retrieved sources, reducing unsupported outputs and improving user trust in delivered answers.
  • Developed a reusable retrieval and evaluation layer that maintained precision across newly ingested datasets, eliminating repeated per-use-case rebuilds and accelerating time-to-production.
  • Deployed the full stack on Kubernetes with CI pipelines and automated rollout, enabling reliable beta releases to customers and repeatable deployments across environments.
  • Instrumented end-to-end metrics and regression tests for retrieval and agent reasoning, enabling continuous evaluation against labeled sets and preventing regression of retrieval quality.
ISRO
India
Software Engineer Intern
Dec 2022 – Apr 2023
Built and stabilized data pipelines and monitoring for operational reporting systems processing telemetry and operational datasets for space systems.
Tech Stack: Python, Structured logging, Automated alerting, Retry logic, Data pipelines
  • Designed and implemented fault-tolerant Python data pipelines from zero processing ~22GB/day, adding retry logic and failure detection that accelerated operational reporting by 40%.
  • Reverse-engineered undocumented production infrastructure and rebuilt components for scale, delivering working systems under ambiguity without a complete specification.
  • Introduced structured logging and automated alerting to surface failures quickly, raising system availability to 99.5% and improving on-call response times.
  • Built retry and backoff strategies and end-to-end validation checks that reduced data loss and improved pipeline correctness in edge failure scenarios.
  • Documented architecture and operational runbooks for the team, enabling smooth handoff and reducing incident resolution time during shift changes.
  • Collaborated with stakeholders to translate operational requirements into measurable pipeline SLAs and monitoring dashboards used for daily operations.

Projects

LLM Evaluation and Benchmarking Platform - SMU AI Research | Jan 2025 – Jun 2025
Tools Used: Ollama, Kubernetes, Terraform, Ray, MLflow, React, Next.js
  • Built an evaluation pipeline to benchmark locally-hosted open-weight models (Ollama) on task completion, consistency, relevance, and output-structure adherence.
  • Provisioned distributed evaluation infrastructure across AWS and GCP using Kubernetes and Terraform; parallelized runs with Ray and tracked experiments in MLflow, cutting experiment-to-deployment time by 30%.
  • Shipped a React/Next.js dashboard surfacing evaluation metrics and pipeline health, replacing manual tracking and improving measured model selection accuracy by 12%.
IC Mapper - Multi-Agent Retrieval System (predecessor to Faradyne) | Aug 2025 – Dec 2025
Tools Used: LangGraph, Qdrant, Cross-encoder reranking, React, Next.js
  • Architected a multi-agent LangGraph system with tool-use function calling and state management to serve complex document corpora for beta users.
  • Diagnosed agent failures as retrieval errors and added sparse retrieval alongside dense embeddings with cross-encoder reranking, reducing agent failure rate by 35%.
  • Delivered an operator dashboard (React/Next.js) exposing agent reasoning and retrieval decisions to enable production debugging under real user load without dedicated ops support.
LlamaFarm - Open-Source LLM Runtime Tooling | Nov 2025 – Feb 2026
Tools Used: Go, Python, Docker, CI, CD, CLI tooling
  • Contributed two upstream-merged PRs to a Go codebase adding a REST endpoint, Python proxy layer, and CLI subcommand exposing real-time model state across engines.
  • Engineered a 5-platform binary distribution resolver with retry/backoff to mitigate cold-start reliability for local model deployments.
  • Built a Docker-based CI/CD test matrix across platforms to produce a repeatable contribution pattern adopted by other maintainers.

Education

Southern Methodist University
M.S. Computer Science (AI & ML)
Dharmsinh Desai University
B.Tech Computer Science

Certifications

AWS Certified Solutions Architect – Associate
HashiCorp Certified: Terraform Associate
Databricks Certified Generative AI Engineer – Associate
Databricks Certified Data Engineer – Associate
Google Data Analytics Professional Certificate

Powered by Drivetube · Create your own profile at drivetube.ai