SaaS
From Telegram feeds to ranked shortlists — ~80% less manual IT recruiting
Approximately 80% less recruiter time on vacancy intake and shortlisting (estimated); ~86% lower monthly sourcing cost versus job-board subscriptions plus manual Telegram monitoring; production go-live in ~20 weeks.
20 wk
go live
Yes
multi tenant
~86%
cost reduction
~80%
manual reduction
The problem
European IT staffing teams sourced a large share of open roles from Telegram channels — but recruiters spent hours scrolling unstructured posts, copying details into spreadsheets, and manually comparing candidates. Duplicate listings, missed roles, and slow shortlists made scaling dependent on hiring more people. Mainstream ATS tools covered job boards, not Telegram sourcing or explainable AI matching against an in-house candidate pool.
Solution
IT grows designed and shipped a multi-tenant recruitment automation platform spanning ingestion, matching, and recruiter workflow. Per-tenant Telethon workers ingest configured channels on a Celery schedule, parse and deduplicate vacancies, and normalize salaries, grades, and work format. Rule-based classification tags roles; Groq-powered LLM scoring ranks candidate–vacancy fit with transparent rationales. Recruiters work in a React web app — moderation inbox, applications kanban, candidate profiles with resume upload, and isolated workspaces per agency. The stack runs containerized on PostgreSQL, Redis, and FastAPI, deployed to production via GitLab CI and Kubernetes.
Architecture
How components connect in production — simplified for clarity.
Sources
Systems & docs
AI layer
Agents / LLM
Approval
Human guardrails
Actions
Integrations
Monitor
Metrics
Results
Approximately 80% less recruiter time on vacancy intake and shortlisting (estimated); ~86% lower monthly sourcing cost versus job-board subscriptions plus manual Telegram monitoring; production go-live in ~20 weeks.
20 wk
go live
Yes
multi tenant
~86%
cost reduction
~80%
manual reduction
Proof from this sector
Real projects with measurable outcomes — not generic promises.
SaaS
4 AI engineers embedded in 12 days
40% faster feature delivery without local hiring overhead.
SaaS
FAQ-grounded AI chatbot — ~60% first-line deflection with one-script embed
Approximately 60% first-line deflection on typical sales and support questions (estimated); one-script website embed in about one day; 170+ FAQ items across three languages with audit-ready LLM usage per organization.
Your industry
We'll share a tailored path for your sector — automation, engineers, or both.
Talk to an AI & Delivery Expert
30 minutes · No commitment
From idea to AI-powered delivery in weeks, not months.
Let's discuss how AI and engineering can accelerate your business.
Automation, engineers, or both — we'll recommend a clear path and respond within one business day.
Talk to an AI & Delivery Expert