AI Internship Batch 1
Build a production-grade AI agent from your own job — in 12 weeks.
12 September – 12 December 2026 · Karachi
An application-based, proof-of-work residency for working professionals. Bring a real problem from your work, design it as an AI agent with PADDLE, and ship it with the batch — evaluated, secured, deployed and defended on demo day.
A private AI community and residency in Karachi. Professionals keep their own domain — finance, health, logistics, education, operations — and learn to embed intelligent systems into it, from first prototype to governed deployment.
Proof of work
Admission by review. A pull request every Friday, a mentor review every Sunday.
12 weeks, 8–10 hours a week
Week 0 kickoff, 12 working weeks, one public demo day.
NED University, Karachi
One in-person lab every Saturday at HPCC; the rest is virtual.
Verified certificate
Human-graded, verifiable online; open-source capstone with your name on it.
National Center in Big Data & Cloud Computing (NCBC), NED University of Engineering & Technology, KarachiWhat the name means
AQLENTIS — a coined word with a real root.
عقل
Aql · intellect, reason, judgement
A coined word with a real root. عقل — aql — is intellect, reason and judgement. The rest expands into what the network does.
Pronounced ak-LEN-tis. The AQL root is written in ember throughout the identity; ENTIS carries the Latin ending for “a being” — the reasoning entity.
AI
artificial intelligence, applied inside real domains
Quotient
the AI Quotient we measure at entry and exit
Learning
structured, mentor-reviewed, proof-of-work
Enablement
professionals enabled, not replaced
Network
a private community of residents, alumni and mentors
Training
hands-on, in the lab, on real problems
Intelligent
Systems
agentic, retrieval-grounded, evaluated, deployed
Your domain × AI = Transformation
Every resident travels the same three stages.
The starting point is always their own profession, never a generic toolkit.
ہنر
Hunar skill
A domain audit. We map what you already know better than any model — the workflows, the exceptions, the data nobody has cleaned — and pick one problem worth solving.
راستہ
Raasta path
AI integration. Retrieval, agents, tool use, evaluation and guardrails, taught on your problem and reviewed weekly by mentors who ship these systems for a living.
منزل
Manzil destination
A deployed use case. Tested, observable, documented, and defended at demo day — with a verifiable credential that links every skill claim to real evidence.
The route
Every Saturday a lab, every Friday a pull request.
Design your agent on paper with PADDLE, then build and ship it.
Week 0
Sat 12 Sep
Kickoff and use-case selection
Baseline assessment, PADDLE walkthrough, every resident pitches a use-case, one capstone chosen.
Weeks 1–2
19 and 26 Sep
Foundations
How LLMs work, validated JSON outputs, pull-request workflow, a tested FastAPI service.
Weeks 3–4
3 and 10 Oct
Retrieval
Corpus, chunking and embeddings in pgvector, hybrid search, grounded answers with citations.
Weeks 5–6
17 and 24 Oct
Agents
Tool calling, LangGraph state machines, MCP servers, memory, approval gates, guardrails.
Weeks 7–8
31 Oct and 7 Nov
LLMOps, security and governance
LangSmith evals, prompt versioning, cost budgets, OWASP LLM Top 10, threat model.
Weeks 9–10
14 and 21 Nov
Capstone sprints
Three subsystems wired end to end, CI green in Docker, thin UI, public staging URL, runbook.
Weeks 11–12
28 Nov and 5 Dec
Hardening and defence
Audit and load test, portfolio writing, a 20-minute architecture defence for each pair.
Demo Day
Sat 12 Dec
Public demo and graduation
Live demo for NCBC, NED and employers; certificates issued, Cohort 02 announced.
Design before code
PADDLE Design Methodology
A six-stage AI-agent use-case design template — Problem, Action Space, Data & Knowledge, Decision Logic, Loop & State, Evaluation. One template = one agent, completed on paper before the IDE is opened, then reviewed and scored by a team lead.
Who should apply
- Working professionals from any domain — logistics, finance, data, CRM, operations
- You bring a real problem from your job; we turn it into an agent use-case
- Some coding helps; Week 1 opens with a Python refresher
- 8–10 hours a week, Saturdays in Karachi
What you leave with
- A PADDLE brief for your own use-case, ready for your employer
- Your subsystem in a deployed, open-source capstone
- A GitHub portfolio of traced, evaluated, secured work
- A verified certificate and induction into the AQLENTIS Guild
Every week
- Saturday
- In-person lab at HPCC, NED
- Wed 8–9 PM
- Virtual clinic on Google Meet
- Fri 6 PM
- Deliverable due as a pull request
- Sunday
- Mentor review of every PR
- Mon 9 PM
- Written feedback and scoreboard
Three ways in
One master identity, three programmes.
The residency trains builders; the enterprise programme takes the same model into organisations; Labs is where the research and open source live.
For professionals
AQLENTIS Residency
Twelve weeks, proof of work, mentor reviews and a verifiable credential. Cohort 01 (AI Internship Batch 1) starts 12 September 2026; Cohort 02 applications open after demo day.
- Application-based, mentor-reviewed
- Hybrid: virtual + weekly lab session at NED
- Graduate with a GitHub portfolio, tests, CI and a live demo
For organisations
AQLENTIS Enterprise
Send your people; get back working AI inside your own workflows. Cohorts are built around one company's use cases, with an organisational AIQ measured before and after.
- Executive and workforce tracks
- Use-case discovery to deployed pilot
- Governance, security and evaluation included
For the field
AQLENTIS Labs
Applied research, open-source tooling and the evaluation datasets behind the residency. Everything we learn about running AI in production, published.
- Open evaluation harnesses
- Architecture decision records
- Research notes and case studies
AI Quotient, measured
Entry → exit, on four dimensions.
The Artificial Intelligence Quotient — a person's ability to use AI to perform a wide variety of tasks — is a construct now studied in peer-reviewed research. AQLENTIS assesses it at residency entry and exit across four dimensions, and reports an organisational AIQ before and after every enterprise cohort.
Every skill claim on a credential links to a project, a commit, a review and a test result.
- 01
Domain fluency with AI
entry → exit
- 02
Prompt, retrieval & agent design
entry → exit
- 03
Evaluation & reliability
entry → exit
- 04
Deployment, security & governance
entry → exit
One stack for the whole batch
Trust, stated plainly
Governance is not a footer link.
These are the commitments every resident, partner and client can hold us to.
Human-controlled grading
AI assists reviewers; it never makes a final admission, removal or certification decision.
Verifiable credentials
Every certificate carries an ID and a public verification page. Revocations are visible.
Terms published before applications open
Hours, stipend status, project ownership, IP and completion criteria — in writing, first.
Your work is yours
Residents own their portfolios. Community content is never used for external model training without explicit consent.
Least privilege for every agent
No AI agent receives unrestricted credentials; destructive or external actions require human approval.
Accessible by design
WCAG 2.2 AA is the target, tested with real assistive technology, in English and Urdu.
Batch 1 poster
Share it with your employer
Intelligence, engineered.
Learn deeply. Build responsibly. Ship intelligence.