DEV Community

Daniel Romitelli profile picture

Daniel Romitelli

Senior SWE @ Marutai Inc. | AI infrastructure builder | Navy veteran | Builder of Scenematic, Thermodynamic RAG & production multi-agent systems. I write about what I actually ship.

Location East Aurora, NY Joined Joined on  Personal website https://craftedbydaniel.com github website

Education

Self-taught engineer | U.S. Navy Aviation Warfare Systems Operator

Work

Senior Software Engineer at Marutai Inc. | CEO at Cloudelipute LLC

The Signal-Processing Boundary That Keeps Coaching Useful in Real Time

The Signal-Processing Boundary That Keeps Coaching Useful in Real Time

Comments
7 min read

Want to connect with Daniel Romitelli?

Create an account to connect with Daniel Romitelli. You can also sign in below to proceed if you already have an account.

Already have an account? Sign in
Fresh Enough to Render: How I Encode Market-Data Trust in the Cache Layer

Fresh Enough to Render: How I Encode Market-Data Trust in the Cache Layer

Comments
8 min read
Text in a Frame Is Contamination, Not Decoration

Text in a Frame Is Contamination, Not Decoration

Comments
10 min read
I Gave My Video Generator Scratch Paper — How Think Frames Saved My GPU Budget

I Gave My Video Generator Scratch Paper — How Think Frames Saved My GPU Budget

Comments
8 min read
The Boundary That Makes iOS Capture Safe on the Web

The Boundary That Makes iOS Capture Safe on the Web

1
Comments
7 min read
How I Stopped Empty Tray Captures From Reaching Whisper in Yapper

How I Stopped Empty Tray Captures From Reaching Whisper in Yapper

Comments
7 min read
How I Built a Patient Check-In Kiosk for a Specialty Medical Practice

How I Built a Patient Check-In Kiosk for a Specialty Medical Practice

Comments
13 min read
Caching LLM Extractions Without Lying: Conformal Gates + a Reasoning Budget Allocator

Caching LLM Extractions Without Lying: Conformal Gates + a Reasoning Budget Allocator

Comments
9 min read
The Day My AI Forgot Everything (So I Built a Context-Continuity Inference Stack)

The Day My AI Forgot Everything (So I Built a Context-Continuity Inference Stack)

1
Comments
14 min read
I Stopped Letting Emails Poison My Extractor: The Pre-LLM Gate That Made the Rest of the Pipeline Reliable

I Stopped Letting Emails Poison My Extractor: The Pre-LLM Gate That Made the Rest of the Pipeline Reliable

Comments
12 min read
I Turned Temperature Up to Save My Extractions: The 3‑Node LangGraph That Trades Variance for Truth

I Turned Temperature Up to Save My Extractions: The 3‑Node LangGraph That Trades Variance for Truth

Comments
12 min read
The Six‑Tier Enrichment Cascade: How I Stop “Helpful” Data From Overwriting True Data

The Six‑Tier Enrichment Cascade: How I Stop “Helpful” Data From Overwriting True Data

Comments
8 min read
User Corrections Always Win: The Streaming Outlook Add‑in UI That Turns Human Edits Into Training Signal (Series Part 4)

User Corrections Always Win: The Streaming Outlook Add‑in UI That Turns Human Edits Into Training Signal (Series Part 4)

Comments
7 min read
I Hardcoded the Kill Switch: Feature Flags as AI Guardrails (Series Part 5)

I Hardcoded the Kill Switch: Feature Flags as AI Guardrails (Series Part 5)

Comments
12 min read
The Queue Was a Table: How I Built Claim/Unclaim Workers with SKIP LOCKED, Stale Recovery, and Retry Caps

The Queue Was a Table: How I Built Claim/Unclaim Workers with SKIP LOCKED, Stale Recovery, and Retry Caps

2
Comments 1
12 min read
The CRM Sync Engine I Had to Reverse‑Engineer: Two‑Step Fetches, 50‑Field Limits, and a Mapper That Refuses to Drift

The CRM Sync Engine I Had to Reverse‑Engineer: Two‑Step Fetches, 50‑Field Limits, and a Mapper That Refuses to Drift

Comments
12 min read
The Search Router That Saved Me From One Index to Rule Them All: Azure AI Search for CRM, pgvector for Transcripts, the CRM a...

The Search Router That Saved Me From One Index to Rule Them All: Azure AI Search for CRM, pgvector for Transcripts, the CRM a...

1
Comments
10 min read
I Hid the AI’s “Thinking” in Plain Sight: Dual-Channel Streaming for an AI Search Chatbot That Works Mid-Call (Series Part 9)

I Hid the AI’s “Thinking” in Plain Sight: Dual-Channel Streaming for an AI Search Chatbot That Works Mid-Call (Series Part 9)

1
Comments
7 min read
I Almost Built a Second Search Index—Then I Realized Privacy Was a Runtime Toggle

I Almost Built a Second Search Index—Then I Realized Privacy Was a Runtime Toggle

1
Comments
8 min read
I Bought the Cheapest Redis and Dared It to Fail: The Circuit Breaker That Made Cache Optional (Series Part 11)

I Bought the Cheapest Redis and Dared It to Fail: The Circuit Breaker That Made Cache Optional (Series Part 11)

1
Comments
8 min read
I Lost Three Hours to a Blank Slate—So I Made “Forgetting” Structurally Impossible (Series Part 12)

I Lost Three Hours to a Blank Slate—So I Made “Forgetting” Structurally Impossible (Series Part 12)

1
Comments
9 min read
Firecrawl Part 2: The Confidence Gate That Decides When Bing Gets a Vote

Firecrawl Part 2: The Confidence Gate That Decides When Bing Gets a Vote

1
Comments
8 min read
Tracing an Extraction Pipeline Like a Ledger: Trace Nodes, DLQ Boundaries, and Replayable Failures

Tracing an Extraction Pipeline Like a Ledger: Trace Nodes, DLQ Boundaries, and Replayable Failures

Comments
10 min read
Per‑Region PBR From One Photo: The Cropping Trick That Stops RGB‑X From Bleeding Materials Across Boundaries

Per‑Region PBR From One Photo: The Cropping Trick That Stops RGB‑X From Bleeding Materials Across Boundaries

Comments
8 min read
My Three‑Phase Parallel Orchestrator: Typed Results, Exception‑Proof Phases, and a Rollout That Never Flaps

My Three‑Phase Parallel Orchestrator: Typed Results, Exception‑Proof Phases, and a Rollout That Never Flaps

1
Comments
9 min read
The Closed‑Loop Consistency Trick: Keeping Scene 12 Faithful to Scene 1 Without Global Memory

The Closed‑Loop Consistency Trick: Keeping Scene 12 Faithful to Scene 1 Without Global Memory

Comments
10 min read
Search That Refuses to Think: The Pattern‑First Query Parser I Use for Fast Intent + Entity Extraction

Search That Refuses to Think: The Pattern‑First Query Parser I Use for Fast Intent + Entity Extraction

Comments
13 min read
Multi‑Vector Embeddings in Production: Typed Vectors, Cache Keys, and a Generator That Refuses Poison Records

Multi‑Vector Embeddings in Production: Typed Vectors, Cache Keys, and a Generator That Refuses Poison Records

Comments
10 min read
MR‑GRPO in Practice: The Reward Mixer That Stops CLIP From Lying to Your Scene Compiler

MR‑GRPO in Practice: The Reward Mixer That Stops CLIP From Lying to Your Scene Compiler

Comments
8 min read
My Voice Router That Refuses to Think: Pattern‑First Multi‑Agent Orchestration for Sub‑Second Latency

My Voice Router That Refuses to Think: Pattern‑First Multi‑Agent Orchestration for Sub‑Second Latency

Comments
12 min read
Turning CRM Audit Noise into a Transition Graph: Normalizing Events, Sessionizing Creation Bursts, and Extracting Time‑Weight...

Turning CRM Audit Noise into a Transition Graph: Normalizing Events, Sessionizing Creation Bursts, and Extracting Time‑Weight...

1
Comments
8 min read
My RAG Stack for Code Retrieval: pgvector HNSW + Metadata Filters + Reranking (and the Parts I Refuse to Guess About)

My RAG Stack for Code Retrieval: pgvector HNSW + Metadata Filters + Reranking (and the Parts I Refuse to Guess About)

Comments
5 min read
Phase 2 Calibration: Per‑Category OOD Thresholds + Group‑Relative Reward Normalization in My Scene Compiler

Phase 2 Calibration: Per‑Category OOD Thresholds + Group‑Relative Reward Normalization in My Scene Compiler

Comments
10 min read
Notification Adjudication in My Ops Intelligence Agent: Canonical Events, Cheap Arbitration, and a Sender That Refuses to Spam

Notification Adjudication in My Ops Intelligence Agent: Canonical Events, Cheap Arbitration, and a Sender That Refuses to Spam

Comments
9 min read
Adaptive Keyframe Sampling: How I Spend a Frame Budget Like It’s Cash

Adaptive Keyframe Sampling: How I Spend a Frame Budget Like It’s Cash

Comments
11 min read
Cache-First Geocoding with Azure Maps: Key Topology, TTL Heuristics, and Quota Smoothing

Cache-First Geocoding with Azure Maps: Key Topology, TTL Heuristics, and Quota Smoothing

Comments
13 min read
Multi‑Agent Firecrawl Research: My Fallback Chain That Refuses to Pretend It Knows the Company

Multi‑Agent Firecrawl Research: My Fallback Chain That Refuses to Pretend It Knows the Company

1
Comments 3
8 min read
Defensive Multi‑Agent Scoring: How I Made LLM Reviews Clamp, Stream, and Fail Loudly

Defensive Multi‑Agent Scoring: How I Made LLM Reviews Clamp, Stream, and Fail Loudly

Comments
6 min read
Diversification After Scoring: The Step That Stops My Scene Compiler From Picking Five Paraphrases

Diversification After Scoring: The Step That Stops My Scene Compiler From Picking Five Paraphrases

4
Comments
6 min read
I got SAM3 video tracking wrong: the session wasn’t the problem—my reprojection was

I got SAM3 video tracking wrong: the session wasn’t the problem—my reprojection was

Comments
7 min read
MR‑GRPO in Practice: The Reward Mixer That Stops CLIP From Lying to Your Scene Compiler

MR‑GRPO in Practice: The Reward Mixer That Stops CLIP From Lying to Your Scene Compiler

Comments
8 min read
Multi‑Vector Embeddings in Production: Typed Vectors, Cache Keys, and a Generator That Refuses Poison Records

Multi‑Vector Embeddings in Production: Typed Vectors, Cache Keys, and a Generator That Refuses Poison Records

Comments
9 min read
loading...