--- title: "Cognilium AI" headline: "Cognilium AI: revenue, AI products and the team" description: "At Cognilium AI, Ali Ahmed built the client-acquisition engine, led AI products from Paralegent AI to VectorHire, and set up the team’s policies, hiring and admin." canonical_url: https://alioahmed.com/work/cognilium type: case-study role: "AI Solutions Engineer" date_label: "September 2024 – present" status: production date_modified: 2026-10-24 author: name: "Ali Ahmed" url: https://alioahmed.com/ orcid: https://orcid.org/0009-0007-4265-3295 key_figures: - "USD 50K+ client work brought in (April 2026)" - "30+ outbound mailboxes across 3 providers" - "4 AI products led" - "24 sections in the staff policy manual I wrote" entities: - name: "Cognilium AI" type: Organization url: https://cognilium.ai - name: "Upwork" type: Organization url: https://www.upwork.com/ same_as: - https://www.wikidata.org/wiki/Q3344170 - name: "LinkedIn" type: Organization url: https://www.linkedin.com/ same_as: - https://www.wikidata.org/wiki/Q213660 - name: "Microsoft Dynamics 365" type: SoftwareApplication same_as: - https://www.wikidata.org/wiki/Q27556059 - name: "Odoo" type: SoftwareApplication same_as: - https://www.wikidata.org/wiki/Q832300 - name: "Google Workspace" type: SoftwareApplication same_as: - https://www.wikidata.org/wiki/Q509284 person_same_as: - https://www.linkedin.com/in/alioahmed/ - https://github.com/alioahmed - https://orcid.org/0009-0007-4265-3295 - https://x.com/Alioahmed_ - https://about.me/alioahmed - https://dev.to/alioahmed - https://medium.com/@alioahmed - https://hashnode.com/@alioahmed - https://substack.com/@alioahmed - https://huggingface.co/alioahmed - https://bsky.app/profile/alioahmed.bsky.social - https://mastodon.social/@alioahmed - https://www.kaggle.com/alioahmed - https://www.producthunt.com/@alioahmed - https://www.patreon.com/cw/alioahmed - https://www.goodreads.com/user/show/202106301-ali-ahmed - https://fueler.io/alioahmed - https://gitlab.com/alioahmed - https://linktr.ee/alioahmed - https://topmate.io/alioahmed related: - https://alioahmed.com/work/paralegent - https://alioahmed.com/work/bijli-bachao - https://alioahmed.com/work/wonder-women - https://alioahmed.com/work/startup-ecosystem full_corpus: https://alioahmed.com/llms-full.txt site_index: https://alioahmed.com/llms.txt --- # Cognilium AI > Since September 2024 I have been an AI Solutions Engineer at Cognilium AI, an AI engineering firm in Lahore. I built and ran its client acquisition — Upwork, outbound email and LinkedIn — and brought in more than USD 50,000 of client work. I led AI products from Paralegent AI to VectorHire, and set up the policies, hiring and administration the team runs on. **Canonical URL:** https://alioahmed.com/work/cognilium **Role:** AI Solutions Engineer — September 2024 – present **Summary:** At Cognilium AI, Ali Ahmed built the client-acquisition engine, led AI products from Paralegent AI to VectorHire, and set up the team’s policies, hiring and admin. ## Key figures - **USD 50K+** — client work brought in (April 2026) - **30+** — outbound mailboxes across 3 providers - **4** — AI products led - **24** — sections in the staff policy manual I wrote ### How the revenue engine works. Client acquisition did not exist as a system when I joined. I built it end to end and ran it every day — research, qualification, proposals, outreach and follow-up — and it brought in more than USD 50,000 of client work by April 2026. - Research. How the Upwork marketplace ranks and where the demand is — turned into a 90-day plan, a profile and agency structure, Project Catalog offers and proposal templates. - Qualify. Job alerts within ten minutes of posting and an IN / MAYBE / OUT fit grid, so effort goes to work we can win and deliver. - Propose. A six-step operation — fit, client research, solution research, a short proposal, screening answers, follow-up — drafted with the browser extension I built. The solution research is where each proposal was designed: the architecture, the hard problems most teams miss, and a first phase scoped so the client could start. - Reach out. Outbound email on six domains and 30+ mailboxes I set up, warmed in stages and sent within a capacity I calculated, rotating tired batches out. On LinkedIn, a written strategy — a three-stage sequence, customer segments and a content calendar — with research-led outreach to US decision-makers and 15+ AdTech accounts. In 2026 I also evaluated and set up OXYGEN and HeyReach with approval gates and testing before launch, and worked inside the client-acquisition playbook: one-company research, drafts reviewed by a person, review gates. - Close. Replies matched to how experienced the buyer is, rate conversations taken to a call, short follow-ups that offer something concrete such as a scope document or an architecture sketch, and a clean handover into delivery — on a strategy I wrote down: the six-pillar growth strategy, the multi-channel plan (March 2025) and a 19-section sales playbook for the team. - Measure. Knowing what actually worked: I separated a form being filled in from the notification actually arriving, an AI assistant sending a visitor from one citing us, and something being tracked from a real outcome. ### Ship the AI products. As their product manager I defined the problem, wrote the requirements and workflows, coordinated delivery and explained each product to buyers. The engineering was a team effort. - Paralegent AI (in production): AI contract review inside Microsoft Word, against the customer’s own playbook. I led it from idea to production and present it in its product video. - VectorHire — AI recruitment (built and demonstrated): I wrote the recruitment-dashboard requirements: AI applicant cards, score filters, a three-phase pipeline and bulk actions, with four agents assessing each candidate in parallel. - VORTA — Enterprise knowledge assistant (built and demonstrated): Product leadership and the demo script for an assistant that answers from a company’s own sources and hands over to a support person when it should. - ProspectVox — Voice sales AI (built and demonstrated): Product and go-to-market for research, fit scoring, sequences, AI calls and CRM updates. On its own outbound, it reached a 47% connect rate against a typical 10–15%. ### Decision tools inside the ERP a business already runs. I lead Cognilium AI’s Dynamics 365 track end to end: decision optimizers for pricing, demand and inventory, and warehouse pick paths — plus Quote Rabbit, a quoting tool for Business Central. - Pricing: What to charge, decided from the data already in the ERP. - Demand & inventory: What to stock, and when. - Warehouse pick path: The shortest route through the shelves. - Quote Rabbit: Turns a customer’s request into a priced quote: it reads the request, identifies and matches the products, prices them, and stops at a confidence gate so a person reviews anything uncertain. Demo catalogue: 1,000 products and 3,472 supplier offers. Writing quotes back into Business Central is the next step, not yet built. - Also designed — An Odoo order flow with two cooperating agents; An agentic insight chatbot; Enterprise search over scraped product catalogues; A legal document analyser built on retrieval; The AgentCore demo script; Multi-tenant knowledge bases on Google Vertex AI; Multi-agent orchestration with event sourcing; AI support-ticket triage with personal-data redaction, in phases; Solution architecture for an EU e-invoicing platform (Peppol); A visual image de-duplication pipeline. Requirements and proposed architectures for client engagements. ### How the company explains and sells itself. A firm that sells AI has to say clearly what it does, for whom and at what price. I wrote that down and built the system that publishes it. - What we sell: Defined, named, scoped and priced the service lines and three engagement models. - Who we are: Wrote the company knowledge base, the Master Brand Profile, the ideal customers and the voice. - Where we show up: Set up 20+ company platform profiles, and built the Cognilium.ai and Paralegent.ai websites with their enquiry flows. - How we publish: Worked on the content engine — Markdown, claim checks, a CMS draft and human review before anything goes live. My byline is on new articles from 24 August 2026. ### Set up the team — People - Job descriptions for generative-AI and senior AI engineers - The hiring pipeline: interview booking, application and onboarding forms, a careers inbox - Sourcing, offers, onboarding — and a professional exit - Mentored a business-development intern as Team Lead (September – November 2025) ### Set up the team — Policy & behaviour - Organisational behaviour, written down: the values, conduct and ways of working the team agrees to - A 24-section staff policy manual: values, conduct, anti-harassment, attendance, leave, working from home, compensation reviews, reimbursement, assets, exit, whistleblowing, conflict of interest and IT support - The HR manual outline - A 9-section employee NDA ### Set up the team — Setup & routines - Team routines and how decisions are routed — including when the CEO is unavailable - Tools and accounts: company email, workspace and platforms - Vendors and partners: suppliers, the bank, the co-working space - Performance and reviews: targets, evaluations and compensation reviews ### Set up the team — Finance & compliance - The banking relationship, including fixing a USD/PKR remittance-split problem - A rejected PSEB registration renewal, seen through to resolution - Export-code alignment, invoicing and salary documentation People, policies, routines and money. I set them up and kept them running, including while the CEO was unavailable. Drafting and administration — not a legal or accounting qualification. ### What I chose, and what it cost. - Focus over volume. Chose: Back the October 2024 decision to sell generative AI and data automation, not app development. Cost: We turned away easier app work to become known for one thing. - A person signs every proposal. Chose: The extension drafts; a person reviews and sends. Cost: Fewer proposals per day — in exchange for ones a client reads to the end. - Inbox before speed. Chose: Warm every mailbox in stages and send inside a calculated capacity. Cost: Weeks before full volume, so the domains still work months later. - Claims checked before they publish. Chose: Content passes claim checks and human review before it goes live. Cost: Slower publishing, and no number we cannot stand behind. ### Credit where it belongs. Cognilium AI is Mudassir Marwat’s company. He founded it in 2019 and leads it as Founder & CEO; its clients are the company’s clients. I am part of the team. The products were engineered together; this page shows the parts I did myself. Not claimed here: client names, revenue beyond the figure above, or authorship of every model and agent. ## FAQ ### What does Ali Ahmed do at Cognilium AI? Ali is an AI Solutions Engineer at Cognilium AI (September 2024 – present). He built and runs its client acquisition on Upwork, outbound email and LinkedIn, led AI products including Paralegent AI, leads the Dynamics 365 track, and set up the company’s policies, hiring and administration. ### How much client work did Ali bring in? More than USD 50,000 of client work by April 2026, through the acquisition engine he built and ran. ### Can Ali build a client-acquisition engine for my company? Yes. At Cognilium AI he built one end to end: marketplace research and positioning, fast qualification, a six-step proposal operation with his own drafting tool, outbound email infrastructure, a written LinkedIn outreach strategy, outreach-platform evaluation and measurement of what actually converted. ### Which AI products did Ali work on? Paralegent AI (in production), and VectorHire, VORTA and ProspectVox (built and demonstrated). As their product manager he defined the problems, wrote the requirements and workflows, and coordinated delivery; the engineering was a team effort. ### Did Ali found Cognilium AI? No. Mudassir Marwat founded Cognilium AI in 2019 and is its Founder & CEO. Ali is part of the team. ### What did Ali set up inside the company? A 24-section staff policy manual, a 9-section employee NDA, the hiring and onboarding pipeline, team routines and decision routing, tools and accounts, vendor relationships, performance reviews, and banking and compliance administration. ### Is Ali available for a role or a project? Open to remote-global roles and relocation to the Gulf (UAE/Saudi/Qatar), US, or Australia, plus fractional AI-delivery consulting. Reach him at hello@alioahmed.com. --- Rendered from the structured content at https://alioahmed.com/work/cognilium. Facts trace to the canonical bio; render, don't reword.