Make.com Automation Developer — AI Pipeline Build (SignoSI)

@goranm16

Hi, this fits well with my automation work. The real goal here is not just connecting tools, but making a reliable AI report pipeline that takes a client submission, runs several model APIs, validates the output, generates the PDF, and sends it without manual chasing. I’ve built similar flows involving Make.com-style webhook automation, Airtable/CRM records, LLM API calls, JSON parsing, document/report generation, and transactional email delivery. I would start by matching the provided schema to Airtable and Tally, then build the main Make pipeline in small tested sections before adding validation, approval, monitoring, and SendGrid delivery. The main risk is one API returning bad or delayed JSON, so I’d add clear error handlers, retries, logging, and manual release checks instead of letting bad reports go out. Thanks!

@Nuel24

THE MAIN RISK IN A MULTI-LLM PIPELINE IS NOT RUNNING MULTIPLE MODELS IN PARALLEL, BUT HANDLING FAILURES WHEN ONE API TIMES OUT OR RETURNS INCOMPLETE DATA. When you route a single request through multiple LLMs like OpenAI, Claude, Gemini, and Perplexity, the real challenge is making sure one failure doesn’t break the entire workflow or corrupt the final output. What usually breaks systems like this is missing fallback logic and weak error handling between steps. My focus is building stable execution flows where partial failures don’t stop the full process. • Parallel routing across multiple LLM APIs with controlled execution flow • Error handling and fallback logic so failed requests don’t break the pipeline • Structured data mapping so outputs remain consistent across models • Reliable delivery into tools like Airtable, SendGrid, or document generation systems The goal is simple: every run completes cleanly, even when one or more APIs fail or slow down. Are your LLM APIs already connected in a single workflow, or are you still building the routing structure? Send me a message and I can walk through how I would stabilize the setup.

@UtroffR

Hi there, Utroff here, I run U-Flow Solutions and build AI driven automations and pipelines for real production use. The tricky part on SignoSI is not just calling four LLM APIs, it is keeping Airtable as the source of truth, handling partial failures in parallel branches, and making sure retries, shadow calibration, and the manual approval step never create duplicate emails or reports. My approach: • Configure the Airtable base and 4 Make data stores, set up the Tally webhook, and map all 56+ fields with strict JSON validation. • Build the main 14 module Make scenario with HTTP modules to OpenAI, Anthropic, Gemini and Perplexity, using parallel routes plus routers for error handling and timeouts. • Implement the shadow calibration layer, daily outcome validation, and manual release scenarios so QA data is persisted and traceable. • Wire Documint placeholders to Airtable, then trigger SendGrid templates for acknowledgement and final PDFs. I have built similar multi step AI automation flows using n8n and custom HTTP integrations, including lead enrichment pipelines that fan out to multiple APIs, normalize JSON responses, then store structured records for later review. Q1. Are all four model APIs already provisioned with keys and quotas, or should the scenarios include basic usage metering per provider? Q2. For the 30 minute SLA, do you have a hard timeout per model, or should the pipeline proceed if one model fails and still generate a partial yet clearly annotated report? Q3. How strict is the requirement on the Documint layout matching the guide, can we add small hidden fields for internal debug metadata if needed? If this sounds aligned, happy to jump on a quick call or chat to walk through the spec and lock scope, then break it into milestones like core pipeline first and monitoring layer second.

@alexbpo

Hi! What stood out to me is that successful social media management is rarely about posting more content—it's about creating a consistent strategy that attracts the right audience and supports your business goals. Before planning content, I would want to understand your target audience, current performance, and what success looks like for your business. Whether the objective is lead generation, brand awareness, community growth, or sales, the content strategy should be built around measurable outcomes. I can help with: • Social media strategy and planning • Content calendars and scheduling • Post creation and optimisation • Community management and engagement • Analytics and performance reporting • Organic growth strategies • Cross-platform content management One advantage I bring is access to a broader delivery network through my BPO operations. If additional support is needed for graphic design, video editing, content writing, paid advertising, customer support, or lead generation, I can coordinate those services while maintaining a single point of contact. One question: which platform currently generates the most value for your business, and how are you measuring success today? Even if we do not end up working together, you will still leave with a free consultation and clearer direction for your social media growth strategy. Regards, Alex

@michela542

Hey, I'm Michel. Your questions: Make + HTTP + Airtable example: Built lead scorer. Typeform → OpenAI HTTP → parse JSON → create Airtable record → branch based on score. Parallel routing: Use parallel routers with error handlers on each branch. One fail doesn't kill all. Merge with aggregator before PDF step. For SignoSI: Tally webhook → Airtable → 4 parallel API calls (GPT-4o, Claude, Gemini, Perplexity) → parse JSON → aggregator → Documint PDF (38 placeholders) → SendGrid email. Plus validation + calibration scenarios. I deliver: 5 Make scenarios, Airtable base (56 fields), Documint template, SendGrid, 4 data stores. You designed it. I build it. Clean. Timeline & price from your post ($400 / 10-15 days) works for me. Send the specs. I start immediately. Cheers, Michel

@haseebtariq0

Right in my lane - I build exactly this: fan a submission to several LLMs in parallel, parse each model's JSON, merge into one structured doc, and email it with retries + monitoring. I implement to your spec, no redesign. Build: - Tally webhook -> Airtable record (3 tables/56+ fields) - Main scenario: a router fires GPT-4o, Claude, Gemini, Perplexity in parallel via HTTP modules, each with its own error handler + retry-with-backoff; responses parsed back to Airtable - Merge -> Documint (38+ placeholders) -> SendGrid -> delivered inside the 30-min SLA - Ack email, daily outcome-validation, calibration shadow layer + manual-approval release as separate scenarios + the 4 data stores On re-runs I keep records idempotent (skip processed) and push hard failures to a dead-letter table for replay, so one model timing out never corrupts a run. Proof: Axon Bridge - automations across a multi-phase pipeline with LLM calls, JSON parsing + autonomous error recovery; Convo AI - scheduled multi-model (Claude/GPT) producing structured reports. More: haseeb-tariq-portfolio.vercel.app. I work across n8n and Make - same HTTP/router/error-handler patterns. Q: if one model (e.g. Perplexity) fails or rate-limits, do you want a partial report flagged as such, or hold until all four return? 10-15 days works. - Haseeb