§00RésuméPune, Maharashtra

Hill Patel

AI Engineer with 1.5 years of experience building agentic AI systems in production — specializing in multi-agent orchestration, deterministic tool-use boundaries, and human-in-the-loop governance for enterprise workflows. Currently architecting orchestration infrastructure at CloudRedux, alongside independent open-source work in long-horizon agent recovery (Cairn) and persistent developer memory systems (memex).

hillpatel.0710@gmail.com+91 97232 88714LinkedIn ↗GitHub ↗
Download PDF ↗
§01ExperienceMay 2025 — Present

AI Engineer

May 2026 — Present

CloudRedux · Pune, India

  • Architecting a meta-level harness pattern, generalizing execution, validation, and recovery logic across agentic workflows.
  • Driving loop engineering for multi-agent systems, structuring orchestrator-critique-executor cycles with human-in-the-loop control.
  • Designed PRISM’s read-only agent boundary using a 42-entry default-DENY tool allowlist, gating all write actions behind deterministic Python.
  • Automated 12 scheduled compliance and billing jobs (GST/TDS, escalation, reconciliation) across 3 legal entities and 10+ properties.
  • Contributing to FosterX PaaS orchestration engine (v2) — workload scheduling, service lifecycle management, and multi-tenant deployment abstractions.

AI Engineer Intern

Feb 2025 — Apr 2026

CloudRedux · Pune, India

  • Designed and implemented stateful agentic workflows using Vertex AI and Gemini models for enterprise use cases.
  • Built memory-aware AI systems with structured reasoning, vendor evaluation logic, and human-in-the-loop control.
§02Featured ProjectsOpen-source & production
memex — May 2026

Persistent memory for AI coding agents via MCP. Bitemporal knowledge graph served to Claude Code, Cursor, and Gemini CLI; 12+ MCP tools, published on PyPI and npm. Read the case study →

Cairn — Jun 2026

Recoverable long-horizon AI agents framework. Framework-agnostic recovery harness and live benchmark testing “Checkpoints Are Compactions” via re-grounding recovery. Read the case study →

PRISM — Madhuban · Present

E-invoicing & payment reconciliation platform at CloudRedux. Cron-job pipelines for automated e-invoicing, multi-source ledger matching, exception flagging, and scheduled invoice dispatch at scale.

§03Technical SkillsSix domains
LLMs & Retrieval
RAGHybrid / Vector SearchRe-rankingLangChain / LangGraphMulti-Agent OrchestrationGoogle ADKClaude CodeMemory & Storage ManagementWeaviate
Model Optimization
Quantization (1–8 bit)PruningDistillationONNX RuntimeTensorRTCUDA tuning
ML Engineering
CNNsTransformersMultimodal ModelsEvaluation (AUROC, F1, hit@K, latency, cost)
System Design
Distributed SystemsMicroservices ArchitectureEvent-Driven DesignAsync ProcessingAPI DesignRBAC & Row-Level SecurityScalability & High Availability
Backend & MLOps
FastAPIDockerAPSchedulerCI/CD (GitHub Actions)RedisPostgreSQLNeo4jObservability (OTel, Prometheus)GCPAWSSQL
Languages & Tools
PythonPyTorchTensorFlowNumPyPandasOpenCVGradioGitLinux
§04Education & Certifications2022 — 2026

B.Tech in Artificial Intelligence & Machine Learning

Sept 2022 — May 2026

NIMS University Rajasthan · CGPA 8.422

Certifications & Achievements

  • Published a peer-reviewed research paper: Transforming Urban Solutions for Smart Cities through Crowdsourced Feedback (Mar 2025).
  • Completed the RAG and Agentic AI Professional Certificate — Coursera.
  • Open-source impact: imgshape (8.84K+ downloads) and memex-mcp (5K+ downloads) on PyPI.