Intelligent Data Capture & Schedule-Linking Layer for Infrastructure Project Management: Real-Time Actual Progress Tracking (Planning-to-Execution Bridge)
Oil India Limited · Smart Automation · Software
The semantic matching is the whole submission and it is genuinely interesting work in a field nobody else will enter — build the confidence threshold and reviewer queue from the start, because a silent wrong match is the failure mode that matters.
What it actually is
A big construction project has a detailed plan in Primavera, but what actually happens on site comes back as free-text daily reports, spreadsheets and supervisors talking. Nobody reliably connects 'spool erected' in a site diary to the planned activity called 'Erect Line 24-XX', so the schedule is always days behind reality. The ask is a layer that reads whatever the site produces and matches it back to the right activity ID.
What to build
An ingestion and matching layer that accepts heterogeneous site inputs — free-text daily progress reports, discipline-wise spreadsheets, scanned site diaries and Primavera or MS Project exports — extracts activity-level start and finish events from them, and semantically matches each reported event to the correct L5 or L6 activity ID despite different disciplines describing the same physical work in different words, with a confidence score and a reviewer queue for ambiguous matches, a conversational or voice interface for supervisors to report progress in plain language, and a closed-project knowledge base storing real durations and deviations against planned ones so future schedules can be planned from history.
Smallest thing that wins the room
Feed in a supervisor's free-text line saying a spool was erected on line twenty-four, and watch the system match it to the planned activity ID with a confidence score, update the actual start date in the schedule, and show the variance against baseline appear on the Gantt.
How crowded this one gets
A guess, projected from the 2025 statements — the last year where both the submission counts and the winners were published.
Quieter than 72% of the 226 · #64 of 226 by expected field
A normal-sized field. Your idea has to be good, not miraculous.
Why: company-sponsored statements drew the smallest fields of all.
This is a guess, not a fact
Nobody has published 2026’s numbers yet. This is an analysed estimate from last year’s pattern, so please do not take it as the truth — check the live counter on the SIH portal before you decide anything. The range covers the middle half of likely outcomes, so one statement in two lands outside it. Entry closes at 500 ideas per statement, so no range goes past that — a statement that reaches the cap fills and shuts rather than drawing an unlimited crowd. The model reads only three things a team can see before choosing — software or hardware, the theme, and what kind of body posted it — and those explain about a quarter of the variation in last year’s field sizes (R² 0.25 on held-out statements). Trust the band more than the number, and the ordering more than either. It cannot see how good your idea is, which is the part that actually decides it.
The scores
The number is the shorthand. The line under it is the reason.
Acceptance potential
4/5A quiet gem — project controls is unglamorous enough that almost nobody will pick it, the semantic matching core is real technical work rather than another dashboard, and the pain is instantly recognisable to any judge who has run a capital project.
Feasibility
3/5The semantic matching core is achievable with sentence embeddings over activity descriptions, and Primavera XER and MS Project XML are documented formats you can parse, but no real project's paired reports and schedules are public so you must construct a realistic work breakdown and report set yourself.
Innovation scope
4/5Matching loosely worded field language to structured activity IDs across disciplines that describe the same work differently is a genuinely open problem, and how you handle field granularity being finer than the planned breakdown is entirely your design.
Clarity
4/5The description states the exact failure precisely, gives the worked example of 'spool erected' versus 'Erect Line 24-XX', and enumerates the required inputs and outputs, though it does not say how accurate the matching must be.
Effort
HeavyMulti-format ingestion, the matching engine, the reviewer interface and the schedule write-back are four pieces, and constructing a believable project schedule with matching reports is itself substantial preparation.
Demo-ability
MediumThe match-and-update moment is clear and satisfying, but a judge needs the planning context explained before they understand why linking a sentence to an activity ID matters.
In its favour
- Green flag: The description hands you the exact matching example, so you know precisely what the hard case looks like and can build and demonstrate against it
- Green flag: The semantic matching core is defensible technical work that separates this from the many project dashboards a judge will see
- Green flag: A voice reporting path for supervisors is explicitly invited and is a small addition that addresses the real reason field data is poor
- Green flag: Project controls is dull enough as a title that the field will be genuinely thin
Against it
- Red flag: No paired real-world schedule and site report dataset exists publicly, so you construct both sides yourself and your matching accuracy is measured on data you authored
- Red flag: Field execution is often finer-grained than the plan, so many reports map to a fraction of an activity rather than to one cleanly — handling partial progress is the hard case and teams tend to skip it
- Red flag: A wrong automatic match silently corrupts the schedule, which is worse than no match, so the confidence threshold and reviewer queue are core requirements rather than polish
- Red flag: The description also asks for a closed-project knowledge base feeding future planning, which is a second product that teams will quietly drop
What you will be writing
- Sentence-Transformers activity description matching
- Primavera XER / MS Project XML parsing
- LLM extraction of start/finish events from free text
- Whisper voice reporting interface
- pgvector similarity with confidence thresholds
- React reviewer queue + Gantt variance view
- Project controls
- Semantic matching
- Construction management
Prior art to read before you start
free-text to structured activity linking · construction progress data capture · schedule variance automation
Analysed by Claude Opus. Every score above is a judgment call with its reasoning attached — kindly cross-check this against the official statement on the SIH portal before your team commits to it.