FACE MATCHDelhi Police
Face recognition for one of India's largest police forces — recognises people from camera feeds using Milvus vector search, with bulk watchlist enrollment and live match alerts for operators.
Hi, I'm Rajeev.
I build real-time systems, from .NET to Angular.
Senior software engineer with 5 years of shipping real-time video-analytics systems — .NET Core backends, Angular frontends, RabbitMQ and PostgreSQL / TimescaleDB. Spent 2 years leading teams of 5 and 10 on the CIAL Airport and Jagannath projects and delivered face recognition for Delhi Police.
01 — at a glance
Scale
9.5k+
connected devices on one platform
Throughput
4×
70 → 300+ frames/sec ingestion
Reliability
99.5%
delivery success at 100+ events/sec
Open source
Contributor to Angular and OrchardCore
Merged PR #19891 in the OrchardCore Media module, with three further feature PRs open. In Angular, PR #70763 documents why an installed service worker isn't replaced when only its response headers change — approved by an Angular maintainer and labelled target: patch.
Daily stack
var me = new Engineer {
Stack = ".NET",
UI = "Angular",
Bus = "RabbitMQ",
Db = "Postgres",
Base = "Gurugram"
};
02 — experience
Where I've worked
i2V Systems
Oct 2021 — NowSenior Software Engineer
- Led a 5-person team on CIAL Airport (ICCC) and a 10-person team on Jagannath (ANPR) over 2 years — sprint planning, design and code reviews, QA coordination and client delivery, while hands-on in the codebase.
- Architected backend services for a city-scale platform — 4,000 cameras, 3,000 streetlights, 1,000 underground sensors and 1,500+ tracked vehicles — integrating an external AI vision-inference service with health checks.
- Built 4 platform modules (ANPR, Actions, Logging, Events) and a Redis caching layer that eliminated repeated per-request database lookups and cut hot-path latency.
- Scaled the event-driven ingestion pipeline on RabbitMQ/MQTT from ≈40 to 100+ events/sec at 99.5%+ delivery success by cutting message payload size, with idempotent consumers and ack/nack retries.
- Diagnosed and fixed production performance issues — profiled memory and CPU on deployed systems and removed a single-threaded bottleneck, raising throughput 70 → 300+ frames/sec.
- Built secure REST APIs, gRPC services and Angular apps on a Facade-based layered architecture, including a real-time GIS map with live vehicle tracking and route trails from GPS telemetry.
- Owned the product's Windows and Docker releases, automating builds and packaging with GitHub Actions; products cleared STQC and CDAC security certification (Government of India).
i2V Systems
2021Software Engineering Intern
Shipped .NET Core REST APIs and Angular features across backend and frontend modules.
03 — deployments
Shipped for real-world clients
Systems I built and delivered at i2V for police, public bodies and critical infrastructure — across India and Turkey.
FACE MATCH
FACE MATCHIGR
Person recognition backed by Milvus vector search — enrollment, deduplication and matching, with async progress over SignalR for large batches.
COMMAND CENTRECIAL Airport
Led the airport's Integrated Command & Control Centre — video wall, live event monitoring and real-time alerts, bringing airport cameras and devices into one operator view.
COMMAND CENTRETurkey ICCC
International command-centre rollout — video wall, event watching and alerts, with central node registration and remote resync across nodes.
PLATE READJagannath
Led delivery of automatic number plate recognition — vehicle detection, plate reads, and SOS and other automated actions on detected events.
04 — projects
Things I've built
↗
dotnet-event-pipeline
Event pipeline sustaining 100k+ events/min on a single machine with at-least-once delivery — transactional outbox, PostgreSQL COPY, Channels backpressure and DLQ retries.
↗efcore-dapper-bulk
Hybrid EF Core + Dapper data layer for bulk inserts and upserts with COPY and UNNEST — cut a local 1M-row insert from minutes to 3–8 seconds.
↗Dyuti OS
India-first desktop OS — a reproducible pipeline that builds a bootable live ISO with Indian-language support.
TimescaleDB Ingestion
Per-frame metadata for ~400 cameras. Removed a single-threaded bottleneck: 70 → 300+ frames/sec.
// platform modules behind these deployments
Face Recognition System
Bulk enrollment, watchlist matching and deduplication with live progress over SignalR.
ANPR Module
Number-plate recognition on Clean Architecture with EF Core migrations, caching and versioned APIs.
ICCC Integration Layer
Node registration, remote resync and RabbitMQ connection lifecycle across multi-node deployments.
GIS / Map Module
Live placement and event overlay of cameras and connected devices on interactive maps.
05 — toolbox
What I work with
Backend
C#.NETASP.NET CoreEF CoreDapperREST APIs
Frontend & GIS
AngularTypeScriptOpenLayersGeoServerPostGIS
Messaging & real-time
RabbitMQMQTTSignalRSystem.Threading.Channels
Data
PostgreSQLTimescaleDBSQL ServerMongoDBMilvuspgvector
Testing & observability
xUnitPlaywrightSerilogSeqOpenTelemetryPrometheusGrafana
DevOps & cloud
DockerGitHub ActionsCI/CDLinux & WindowsAWS · personal projects
06 — elsewhere
What Google's AI says about my open-source work
Asked about RAJEEV2510, Google's AI assistant summarises my contributions across Angular, OrchardCore, ng-bootstrap and the wider JavaScript ecosystem. The verified record lives on GitHub.
Open to full-stack .NET + Angular roles
Let's build something reliable.
Hiring for full-stack .NET + Angular, or want to talk real-time systems? My inbox is open.