ProsecuTrack

ProsecuTrack — Find the gap before the court does.

We came to ProsecuTrack through a prosecutor’s office, not a coding challenge.

As students at the National Law School of India University, we interned at a government prosecutor’s office and saw the work behind every court appearance: tracking evidence, following up on forensic reports, comparing witness statements and keeping procedural deadlines in view. Much of this depended on manually working through files under the pressure of a substantial caseload. We saw how easily a gap could remain buried in the paperwork until it became a question in court.

That experience gave us our starting point: what if the prosecutor could see those unanswered questions earlier?

Learning about the Ministry of Home Affairs’ Inter-operable Criminal Justice System (ICJS) initiative helped us imagine where such a tool could fit. Alongside efforts to connect criminal justice institutions and their information, we saw an opportunity for a prosecutor-facing layer that turns case records into a clear picture of what needs attention.

ProsecuTrack is our prototype for that layer. It brings evidence, law, witnesses and timelines into a case-readiness dashboard. Its central feature, the Case Failure Radar, surfaces potential issues: a forensic report not located in the indexed records, electronic-evidence documentation requiring verification, conflicting incident times or an approaching configured statutory threshold. Each finding carries its factual basis, source, legal framework, reasoning and limitations.

The prosecutor can inspect the finding, challenge it and record a decision. The system supports that judgment; it does not determine guilt, predict judicial outcomes or decide whether prosecution should proceed.

We are starting with criminal prosecution because that is where we encountered the problem firsthand. Once we prove our mettle here, we aim to extend the approach beyond criminal law to departments managing similarly demanding case volumes, records and deadlines, including forest and income-tax administration. Each extension would require its own legal rules and workflows.

The current browser-based prototype uses synthetic records and deterministic demonstration logic. It is not connected to ICJS or government databases and does not yet read or extract information from real case files.

How It Works

1. A case becomes a structured record. The case’s offence classification, custody status and relevant dates provide the foundation for review. Different rules require different date anchors: the demonstration’s configured detention clock uses a first-remand date, rather than simply counting from FIR registration.

2. The legal rule becomes inspectable. The system connects the configured offence and punishment category to the relevant framework, such as BNSS §187(3), and displays the deadline calculation. Users can see the assumptions behind the countdown and identify what requires legal verification.

3. Evidence and witnesses are brought into one view. CCTV footage, medical reports, recovery records and witness statements are represented in an organised index, with their sources, status and connections. An Evidence Passport displays provenance, documentation and synthetic chain-of-custody details.

4. The Failure Radar checks for potential gaps. The demonstration logic compares indexed records and timelines against configured expectations. It flags an expected FSL report that has not been located, digital documentation requiring review or incident accounts containing different times. It does not treat a missing upload as proof that evidence does not exist.

5. Every flag explains itself. Opening a finding reveals its fact, source, law, reason, limitation and suggested verification step. The prosecutor can examine both why the issue was raised and what could make the finding wrong.

6. The prosecutor makes the call. Users can confirm, resolve, dismiss or escalate a finding—or challenge its basis. These actions are recorded in a browser-local audit trail. A human decision changes the review status without silently rewriting the source evidence.

7. The case comes together in a brief. At any stage, the prosecutor can generate a printable and downloadable brief bringing together the configured law, statutory clock, evidence status, potential issues, pending verification and recorded human decisions.

ProsecuTrack makes the unanswered questions visible before the court asks them.

AI identifies. Law governs. Humans decide.

Case Failure Radar — Deterministic detection of case-readiness gaps: missing FSL reports, unverified digital evidence, witness timing inconsistencies, and incomplete investigation items — each with a full FACT / SOURCE / LAW / REASON / LIMITATION / ACTION breakdown.

Statutory Clock — Tracks the default-bail investigation deadline under BNSS §187(3), correctly distinguished from §193, with a step-by-step "Why this deadline?" explanation and a "Show system logic" toggle exposing the raw calculation.

Case Readiness Score — A weighted, fully deterministic 0–100 operational readiness score (procedural, evidence, forensic, witness, digital) — explicitly not a guilt, conviction, or acquittal prediction.

Legal Grounding Chain — Visualizes how every finding traces from fact → document → law → precedent → action → human decision.

Evidence Passport — Per-exhibit chain-of-custody visualization with hash, storage, transfer count, and documentation status.

Precedent Intelligence — Searchable synthetic case-law database with relevance scoring, grounding explanations, and a Verified Answer Mode that refuses to answer rather than fabricate authority.

Challenge AI — Every finding can be challenged by the prosecutor with a recorded reason, moving it into human review.

Human Review — Confirm, Resolve, Dismiss, or Escalate any finding — no finding changes status without an explicit human decision.

AI Audit Trail — Full timestamped log of every finding, its legal basis, the AI action, and the human decision.

Case Graph — Interactive network view of how evidence, people, law, and precedent connect within a case.

What-If Simulator — Toggle hypothetical scenarios (e.g. "FSL report received") to see the readiness-score impact, clearly labeled as simulation only.

Prosecutor Brief Generator — One-click printable/downloadable case brief compiling law, evidence, issues, and required decisions.

Responsible AI dashboard — Explicit status board (guilt/conviction prediction disabled, human oversight/override/audit trail enabled) plus governance principles.

Feedback, mainly — specifically from anyone with real prosecution, investigation, or criminal-procedure experience under BNS/BNSS/BSA. The legal framework in this build (the §187(3) default-bail clock, the BSA §61–63 electronic-evidence distinction, the readiness-scoring weights) was built carefully but is still a hackathon-scope simplification, and I'd want practitioner eyes on:

Whether the Failure Radar categories (statutory clock, forensic gaps, digital evidence, witness consistency, investigation completeness) match what actually bites prosecutors in practice, or whether there are bigger gaps the tool is missing entirely. Whether the readiness score's weighting feels meaningfully calibrated or just plausible-looking. Whether the "Why panel" / Legal Grounding Chain gives a prosecutor enough to actually act on, or whether it's explaining the wrong things.

Also open to collaborators who want to extend this — real document ingestion (OCR/parsing actual case files instead of the synthetic dataset), a proper backend for multi-user/multi-case use, or deeper precedent retrieval against an actual case-law corpus rather than the labeled synthetic authorities used here.