01 Monitor
Know what the country is talking about
One live picture of what monitored sources are posting, where it is about, and what is starting to spread, with questions answered in plain language.
Social Media Intelligence Platform
RAVEN
An end-to-end OSINT suite for monitoring social media at national scale, detecting manipulated media, investigating where a claim began, and identifying and mapping persons of interest.
MonitorVerifyIdentify
12
Modules
2
Languages
24/7
Collection
0
Data leaving your boundary
Confidential · For authorized national-security use · Bangladesh · 2026
What it achieves
What a unit can do with RAVEN.
Twelve modules share one backbone, so a signal raised in one surfaces in the others. This is what that adds up to for the unit using them. Modules are entitled per organisation, so a deployment carries only what its mandate covers.
02 Monitor
Catch a story before it breaks
See a rumour, a protest call or a panic while it is still building, with time to respond before it peaks rather than after.
03 Monitor
Never miss what you are watching for
Set a standing watch on a topic, a place or a narrative once, and hear about it the moment it resurfaces.
04 Verify
Call out a fake with evidence
Establish whether an image, a video or a recording was made or altered by a machine, with a report that holds up when it is challenged.
05 Verify
Find where a claim started
Trace a rumour or a recycled photo back to its earliest known appearance, and show exactly how it travelled from there.
06 Identify
Build a subject file in hours, not weeks
Bring together what open sources and authorised data say about a person of interest, with every line sourced and every uncertainty labelled.
07 Govern
Make decisions you can defend
Every detection passes a human analyst, and every decision lands on a record nobody can alter afterwards.
08 Report
Brief leadership in one click
Turn any finding into a branded PDF or DOCX report, ready to go up the chain of command.
Monitor
Watch every source in real time.
Continuous collection across monitored public pages, with post and comment analysis in Bengali and English, geospatial mapping, and a virality model that surfaces a story before it breaks rather than after.
Screen 01
Situational dashboard
The first screen after sign-in. Six KPI cards, the posts taking off in the last three hours, and the map of places the posts are about, each card a way into the module behind it.
- Engagement is a weighted score, not a plain sum, so a share counts for more than a like.
- Every card is a link: Posts opens Monitoring, Active Tracks opens Watchtower, AI Flagged opens Detection.
- Dhaka’s 186 posts include the ones that named only a neighbourhood, such as Mirpur.
Posts
3,086
▲ 8% last 7 days
Engagement
1.2M
reactions + comments + shares
Monitored Pages
12
12 total tracked
Active Tracks
5
watchtower
Recent Events
12
last 7 days
AI Flagged
7
64 analysed
About to take off last 3 hours · judged minutes after posting
CN 87
Channel Nadi TV May go viral
টানা বৃষ্টিতে মিরপুর ১০ ও কাজীপাড়ায় হাঁটুপানি, বেগম রোকেয়া সরণিতে যান চলাচল বন্ধ।
38 min ago 23× → ~61.4K
DD 64
Dhaka Dispatch Gaining traction
Metro rail will run until 11 pm from Saturday, with a train every eight minutes in the evening peak.
52 min ago 4.6× → ~18.2K
BB 52
Bangla Barta 24 Gaining traction
Onion climbs to Tk 140 a kilo at Karwan Bazar; traders expect relief once imports clear the port.
84 min ago 3.1× → ~9.8K
Geographic Activity
Low High
1 Dhaka ঢাকা 186
2 Chittagong চট্টগ্রাম 57
3 Gazipur গাজীপুর 31
4 Sylhet সিলেট 24
5 Cox's Bazar কক্সবাজার 21
Every monitored page, one feed
Continuous collection from the public pages a unit adds, analysed in Bengali and English and searchable in either, with every post linked back to its source.
Continuous CollectionBengali & EnglishBilingual SearchLink to Source
Stories before they break
Posts are grouped into stories and ranked by where engagement is heading, so a surge shows up while it is still starting rather than after it has peaked.
Story GroupingMomentumProjected ReachEarly Warning
A standing watch
Tell Watchtower what to look for once, and it keeps looking, escalating what it finds into the unit’s channels and onto the map.
Standing TracksEscalated EventsGeospatial MappingUnit Alerts
Verify
Tell authentic from synthetic.
Forensic analysis of images, video and audio, paired with an agentic investigator that traces a claim back to its earliest appearance, from a single upload or a URL to a confidence-rated report with its working shown.
Screen 02
AI Detection result
The detector’s verdict and its likelihood. The heatmap marks the region that drove the call, and the indicators say what the model actually saw.
- Recommendations tell an analyst what to check next, so the result starts work rather than ending it.
- Report downloads the branded PDF; the raw API response stays one click away.
- With Facebook selected, the drop zone becomes an address field and the server fetches the media.
AI Generated 91.4% AI likelihood 91.4%
Media Analysis
Original
Manipulation Heatmap
Key Indicators
- water reflections do not match the buildings above them
- rickshaw wheels merge into the water surface
- shop signboard lettering is malformed and unreadable
- uniform noise pattern across sky, walls and water
- overly smooth transitions at object edges
Recommendations
- Look for news coverage of the rainfall on the same date
- Check the signboard text against real shops on the street
- Search for other photos of the same street that day
Synthetic media forensics
Images, video, audio, or a link to a post. The verdict comes with its likelihood, the region that drove the call, and what to check next.
Image · Video · AudioPost URL InputRegion HeatmapBranded PDF
Claim-origin investigation
Root Analysis traces a claim, a picture or a post back to its earliest known appearance, and returns per-claim findings, a timeline and its stated limitations.
Earliest SourcePer-Claim FindingsTimelineWorking Shown
Identify
From open sources to a structured dossier.
Identity Intelligence assembles a subject file from open sources and the data a deployment is authorised to hold, records where every line came from, and labels anything it is not sure about. Photos are read for what is in the frame, and a reading is a lead, not a conclusion.
Structured subject dossiers
Assembled from open sources and the data a deployment is authorised to hold. Every line records where it came from.
Open SourcesAuthorised DataPer-Line ProvenanceRisk Level
Leads, not conclusions
Anything RAVEN is not certain of is labelled as such, and that confidence travels with it into the report.
Confidence MarkedVerify Before ActingExecutive SummaryReport Ready
Casework & disposition
Detections needing a human decision land in one queue. An analyst dispositions each, and who decided what, and when, is written to a record nobody can edit afterwards.
Single QueueHuman DispositionImmutable LogHand-Off
Platform
Built to deploy inside your boundary.
RAVEN runs where the mandate requires it to run. Modules are entitled per organisation, retention is set per organisation, and every screen states plainly what it cannot tell you.
Deployment
- On-premises or sovereign cloud
- Per-organisation module entitlement
- Nothing crosses the deployment boundary
Collection
- Public pages only, added by the unit
- Continuous scheduled collection
- Re-observation of fast-moving posts
Governance
- Role-based access control
- Immutable action log
- Human disposition on every detection
- Retention set per organisation
Transparency
- Counts are snapshots, for ranking not measuring
- Confidence stated on every finding
- Every screen states what it cannot tell you
Authorized access
Built for national-security missions.
RAVEN is made available exclusively to authorized government agencies and national-security institutions. The full walkthrough is given in person: contact us to arrange a briefing and a controlled demonstration.
Confidential · RAVEN Social Media Intelligence Platform · Bangladesh · 2026