Project Selection and Planning
Open in the Padhaiko app2.1 Selection of major project topic based on theory/practical subject studied or as per the research or industry needs
Selecting a Major Project topic is the single most consequential decision in Project I because your group will live with that problem across both the seventh semester (ENCT 413: Project I) and eighth semester (ENCT 461: Project II). A strong topic sits at the intersection of academic depth (computer engineering principles), real-world relevance (research or industry need), and two-semester feasibility.
Sources of a Major Project Topic
Under the ENCT 413 syllabus, a major project topic originates from one of three primary sources—or an extension of the sixth-semester Minor Project—and must align with computer engineering domains:
- Theory or practical subjects studied: Deepening concepts from core courses such as Operating Systems, Database Management Systems, Computer Networks, Artificial Intelligence, Computer Graphics, Embedded Systems, or Software Engineering.
- Research needs: Investigating an open technical challenge—improving accuracy, robustness, latency, or resource efficiency on a defined dataset or simulation benchmark.
- Industry or community needs: Addressing an authentic operational bottleneck identified through field visits, local institutional partnerships, or an internship host organization.
| Syllabus Domain | Theory / Practical Origin | Research-Need Example | Industry / Community-Need Example |
|---|---|---|---|
| AI, ML & Computer Vision | Neural networks, image processing labs | Lightweight Nepali Devanagari license-plate recognition under low-light glare | Automated visual defect screening for a local packaging unit |
| NLP & Audio | Compiler/automata, AI & speech processing | Fine-tuning low-resource Nepali speech-to-text for noisy clinic environments | Bilingual voice-assisted appointment triage for rural health posts |
| Networks & Cybersecurity | Computer networks, cryptography, OS | Anomaly-based DDoS detection using lightweight flow features on edge routers | Automated vulnerability & compliance scanner for campus/SME web portals |
| Embedded Systems & IoT | Microprocessors, embedded systems, instrumentation | Energy-aware LoRa routing and sleep scheduling for landslide sensor nodes | Smart greenhouse micro-climate monitor with offline alert fallback |
| Big Data, Web & Mobile | DBMS, distributed systems, web engineering | Comparative indexing and query latency study on time-series telemetry | Offline-first cooperative inventory and dairy ledger mobile platform |
2.1.1 Evaluating Candidate Topics Before Committing
Before locking in a single idea, shortlist two to three candidate topics and test each one against five practical filters:
- Two-semester runway: Does the problem justify both Phase I (design, dataset/setup, proof of concept) and Phase II (full build, rigorous evaluation, deployment)? Avoid tiny CRUD apps that finish in two weeks, and avoid moonshots requiring supercomputers you cannot access.
- Data and hardware reality: Can you legally and reliably obtain the dataset, sensors, servers, or testbed needed? If a topic depends on hospital MRI scans or paid cloud GPUs you do not have, pivot early.
- Track alignment: Does the topic match your chosen track (Research, Product, or Internship-Based)?
- Supervisor expertise: Is there a faculty supervisor in the department who actively works in or understands this domain?
- Measurable outcome: Can you clearly state how you will prove the project succeeded at the end of Project I and Project II?
2.2 Problem identification and scope definition
Once a domain and candidate topic are chosen, you must turn a broad theme (for example, "Smart Traffic Control" or "Mental Health Chatbot") into a sharp engineering problem statement, measurable objectives, and an explicit scope boundary.
Problem Identification and Scope Definition
- Problem Identification is the process of pinpointing who faces a specific technical or operational problem, where current methods fall short (with evidence from observation, field visits, or literature), and why solving it matters.
- Scope Definition draws an explicit boundary around the system: specifying in-scope deliverables (what will be engineered and evaluated across Project I and II), out-of-scope exclusions, assumptions, and constraints.
2.2.1 Writing a Focused Problem Statement
A crisp problem statement fits in one to two short paragraphs and answers four questions without marketing hype:
- Context & target user: Who operates or is affected by the current process?
- Core deficiency: What specific limitation (error rate, manual delay, high cost, lack of offline support, vulnerability) exists today?
- Evidence of the gap: What did a field visit, sample benchmark, or literature check reveal?
- Engineering consequence: Why does a computer engineering solution fix that bottleneck?
2.2.2 Structuring Objectives and Phased Scope (Project I vs. Project II)
Because the major project spans two semesters, defining scope properly means separating Phase I milestones (Project I) from Phase II completion (Project II) while keeping out-of-scope items explicit to prevent scope creep.
| Scope Component | What to Define in Project I (Sem 7) | What Carries Forward to Project II (Sem 8) |
|---|---|---|
| In-Scope Core | Problem formulation, literature survey, architecture design, dataset/environment setup, baseline algorithm or core module, Proof of Concept (PoC) | Remaining feature modules, full system integration, optimization, ablation/comparative testing, user or field evaluation, deployment |
| Out of Scope | Clearly list features you will not build at all (e.g., iOS native app, multi-currency payment gateway, hardware PCB mass manufacturing) | Remain out of scope unless formally revised with supervisor approval |
| Measurable Objectives | 3–4 SMART objectives covering the complete major project, with Phase I PoC targets flagged | Final verification of all 3–4 SMART objectives against benchmarks |
| Constraints & Assumptions | Team size (max 3), budget, hardware specs, available compute, dataset license/privacy | Same constraints, plus deployment/hosting environment constraints |
2.2.3 Initial Work Planning and Risk Register
After defining scope, build a realistic Project I work plan (typically a 12–14 week schedule aligned with Institute of Engineering milestones: concept note in Week 2, proposal in Week 3, progress report two weeks before semester end) and a short risk register identifying fallback plans if data collection, hardware delivery, or API access slips.
Turning a broad theme into a sharp problem statement and phased scope
A three-member group wants to build a project around "Agriculture using AI" based on their Machine Learning and IoT coursework. Outline how they should narrow the topic, write a problem statement, and define a two-phase scope separating Project I from Project II.
Narrow the broad theme using field relevance and feasibility: "Agriculture using AI" is far too broad. After speaking with local vegetable farmers and an agricultural officer (field inquiry), the group discovers that smallholder tomato farmers misidentify early fungal blights and overuse broad-spectrum pesticides, yet lack reliable mobile internet in greenhouses. They narrow the topic to: Offline Mobile Edge-AI Diagnosis and Treatment Advisory for Tomato Leaf Diseases in Nepali.
Draft a concrete problem statement: Smallholder tomato growers in peri-urban Nepal lose yield to late-detected fungal and bacterial leaf diseases. Existing cloud-based plant-disease apps require continuous internet connectivity, output English-only botanical labels, and drop in accuracy on cluttered field backgrounds. There is a need for an offline-capable mobile diagnostic tool trained on realistic field images that delivers actionable Nepali guidance.
Define phased scope and explicit exclusions:
- Project I (In Scope): Collect and label 1,200 local field + benchmark leaf images across 4 classes; design system architecture; train and quantize a baseline MobileNetV3/YOLO-nano model; build a PoC Android inference screen showing class + confidence.
- Project II (In Scope): Expand dataset, run lighting/background robustness ablation, add Nepali voice/text advisory cards, conduct usability testing with 15 farmers.
- Out of Scope: Automated drone spraying, soil nutrient hardware probes, iOS release, and multi-crop coverage beyond tomatoes.
The group narrows "Agriculture using AI" to offline Nepali tomato leaf disease diagnosis, grounds the problem statement in field constraints (offline need, background clutter, language barrier), and separates Project I (dataset + baseline quantized PoC) from Project II (hardening, advisory UI, and field trials) with clear out-of-scope exclusions.
2.3 Approval of project by the department and prospective mentor in case of internship based track
Topic selection and planning are incomplete until the project receives formal approval from the department (and, for the internship-based track, written endorsement from the prospective industry mentor). The department project committee verifies academic standard, non-duplication, feasibility, and supervision coverage before students invest weeks into development.
Department and Prospective Mentor Approval
Project Approval is the official authorization by the Department Project Committee (typically via Concept Note and Proposal evaluation) confirming that the problem has sufficient B.E. engineering depth, fits the 2-semester timeline, and has assigned supervision. For the internship-based track, formal approval by a prospective mentor at the host organization is additionally required to validate workplace feasibility, domain mentorship, and resource access.
| Track | Internal Reviewer | External Reviewer | What Approval Verifies |
|---|---|---|---|
| Research Track | Assigned supervisor + Department committee | None (unless external lab collaboration) | Clear research question, accessible dataset/compute, non-trivial experimental plan |
| Product Track | Assigned supervisor + Department committee | Optional domain stakeholder feedback | Clear user pain point, sound software/hardware architecture, realistic phased scope |
| Internship-Based Track | Assigned supervisor + Department committee | Prospective industry mentor | Authentic engineering contribution by all group members, NDA/data compliance, mentor commitment |
2.3.1 Preparing for Successful Approval
To avoid repeated rejections or delays during Weeks 2–3:
- Pitch 2 options initially to your supervisor: Briefly compare your top idea and backup idea so your supervisor helps pick the stronger engineering path.
- Show evidence of feasibility: Attach a sample dataset link, hardware bill of materials, or preliminary field-visit notes rather than vague promises.
- For internship track, secure mentor endorsement early: Request an official internship/project confirmation letter stating the organization name, mentor's name and designation, project problem statement, student roles, and confirmation that academic milestones (reports and presentations) are permitted.
- Respond constructively to committee revisions: If the committee asks to narrow scope or switch algorithms, update the scope document immediately and confirm with your supervisor.
Securing Dual Approval for an Internship-Based Project
A group of two students is interning at an energy analytics startup. Their company manager wants them to "help maintain the company's React dashboard and write unit tests." Explain why this would fail department approval and show how to reshape and obtain approval from both the prospective mentor and the department.
Diagnose why maintenance tasks fail academic approval: Routine bug-fixing, styling tweaks, and writing unit tests on an existing commercial app lack an independent engineering problem statement, research/design methodology, and measurable Phase I/II progression required for a 2-credit Major Project.
Negotiate a bounded, self-contained engineering module with the prospective mentor:
With the mentor, they scope a dedicated module: Automated Anomaly and Theft Detection Pipeline for Smart-Meter Load Profiles, exposing a forecasting/anomaly API and visualization view integrated with the company dashboard.
Prepare the dual approval package:
- Prospective Mentor Endorsement: Signed letter/form specifying the problem statement, anonymized historical smart-meter CSV sample access, weekly mentor review hours, and permission to present architectural diagrams and anonymized metrics at Institute of Engineering.
- Department Submission: Clear separation showing what the existing startup platform already had versus what the 2-student team is designing and implementing from scratch (time-series preprocessing, anomaly detection model comparison, API service, and validation metrics).
Routine maintenance is rejected because it lacks original design and methodology. The students negotiate a dedicated, self-contained anomaly-detection module with their industry mentor, obtain signed mentor approval confirming anonymized data access and academic presentation rights, and clearly distinguish their new engineering work from the host's existing codebase for department approval.
Viva-Style Questions: Project Selection, Scope, and Approval
What are the valid sources for selecting a Major Project topic according to the ENCT 413 syllabus? Explain why choosing a technology stack first (e.g., "we want to use Blockchain") before identifying a problem is poor engineering practice.
Syllabus topic sources: Topics may come from (1) theory or practical subjects studied in the B.E. curriculum, (2) research needs, (3) industry or real-world needs (including field visits or internships), or (4) extending the sixth-semester Minor Project.
Problem-first vs. technology-first: Starting with a buzzword technology forces an artificial problem onto a tool that may be unnecessary (for instance, using blockchain where a standard relational database with audit logs is faster, cheaper, and simpler). Good engineering identifies a real bottleneck first, then selects the simplest effective architecture and tools to solve it.
Topics come from studied theory/practical subjects, research needs, industry/community needs, or Minor Project extensions. Picking a technology before identifying a problem leads to over-engineered, poorly justified systems; examiners always ask why the chosen method is needed for that specific problem.
During your proposal viva, an examiner asks: "What is the difference between your problem statement and your scope?" Give a crisp, structured answer with a brief example.
Distinction:
- The problem statement defines the existing pain point, who experiences it, and why current solutions fail.
- The scope defines the exact boundary of your proposed solution—what features, data, and experiments are in scope (across Project I and II) and what is explicitly out of scope.
Example: In a campus bus tracking project, the problem statement is that students waste 20–30 minutes waiting because arrival times vary with traffic and no real-time position feed exists. The scope specifies in-scope work (GPS tracking on 5 campus buses, ETA prediction, Android student view) and out-of-scope exclusions (ticketing, payment integration, external city routes).
The problem statement describes what is wrong today and why it matters, whereas the scope defines exactly what our system will and will not deliver to address it within our time and resource constraints.
Why must a Major Project team separate in-scope deliverables into Project I and Project II from the beginning? What belongs in each phase?
Reason for phased planning: Project I and Project II are evaluated separately at the end of Semester 7 and Semester 8. Without phased boundaries, teams either try to rush full deployment in Semester 7 without solid design, or drift in Semester 7 and panic in Semester 8.
Phase allocation:
- Project I (Sem 7): Problem formulation, literature review, concept note, proposal, data/environment setup, system design, core algorithm/methodology development, and a working Proof of Concept (PoC).
- Project II (Sem 8): Full integration, edge-case hardening, thorough comparative/user testing, deployment, and final thesis report and defence.
Phased scoping aligns deliverables with the two-semester assessment structure: Project I completes research, design, data/setup, and a validated Proof of Concept, while Project II completes full integration, comprehensive evaluation, and final deployment.
Why does the department require approval from both the department committee and a prospective mentor in the internship-based track?
Role of the prospective industry mentor: The mentor verifies that the proposed problem is real, that the organization will provide necessary workspace, guidance, and permissible sample/anonymized data, and that the student's work is feasible within the internship period.
Role of the department committee: The committee verifies that the work meets B.E. Computer Engineering academic standards, is not mere clerical/maintenance work, credits each group member fairly, and complies with Institute of Engineering submission and presentation schedules.
Dual approval protects both sides: the prospective mentor confirms workplace feasibility, mentorship, and data access, while the department committee confirms academic rigor, fair group contribution, and alignment with Institute of Engineering milestones.
Suppose your group proposes an AI-based medical diagnosis system using hospital X-ray records, and the committee raises feasibility concerns during project selection. What concrete adjustments can rescue the project before final approval?
Identify the feasibility bottleneck: Hospital data requires ethical clearance and months of paperwork, which often stalls a 7th-semester timeline before Project I ends.
Concrete scope and data adjustments:
- Replace inaccessible local hospital records with a publicly available, peer-reviewed clinical dataset (e.g., NIH ChestX-ray, CheXpert, or local open-access benchmarks) for baseline training and reproducibility.
- Narrow the clinical claim from "autonomous medical diagnosis" to "computer-aided screening / triage decision support" to stay ethically and technically realistic.
- Confirm access to free/campus GPU resources (or lightweight models trainable on available hardware) and validate the plan with the supervisor before resubmitting.
The team can rescue feasibility by pivoting to validated public medical datasets (with optional small anonymized validation later if clearance arrives), reframing the system as decision-support screening rather than autonomous diagnosis, and verifying compute resources for baseline training.