GET
/profile
→ 200 OK
Engineering Manager · Enterprise Product Platform · API Integrations
Hari Prasath Raman
Everything a client sends to Bloomberg passes through something my team owns. I believe a system you can't see is a system you don't own — so here's the pipeline, and the career that runs through it, instrumented.
Twelve years across trading systems, middleware and network automation. People-first leadership, and a firm bet on AI-assisted engineering as a quality lever, not just a speed one.
Data ingestion pipeline overview
Ingestion platform
Validate · translate · route
DataMapper dictionary · OpenAPI contracts · tracing on every hop
300 msg/s · latency −50%
Bus
Kafka
RabbitMQ
Pub/Sub · SSE
Consumers
Buyside workflows
Sellside workflows
Client subscribers
GET
/experience
?order=desc
Returns the trace first, then each span with its impact as measured outcomes.
- scope
- Own the unified FIX / REST / BLPAPI ingestion platform behind the firm's mission-critical client data workflows — enterprise standards of correctness, reliability and performance.
- impact[0]
- Re-architected trades/orders ingestion: 5 → 300 msg/s, 50 → 1,000+ orders per basket, 1M+ transactions a day across buyside and sellside workflows.
- impact[1]
- Made the platform legible — dashboards and interfaces surfacing ingestion flow, latency patterns, error hotspots and per-client activity in real time; confident operations through peak market events.
- impact[2]
- Drove a shared Pub/Sub platform capability out of competing SSE APIs — OpenAPI and Kafka contracts agreed before implementation so consumers could validate early.
- impact[3]
- Directed the Linux migration at DR-1 with 99.9%+ uptime and no disruption to trading days; latency halved; releases monthly → weekly, error rate −30%.
- impact[4]
- Coach the team on LLM-based tooling for troubleshooting, code assistance, test generation and pattern detection.
- team
- 10 engineers · onboarding ramp 3 months → 1 · TDD/BDD default · three senior engineers now lead areas independently
- impact[0]
- Designed DataMapper, a configuration-driven translation framework serving 600+ requests per second with horizontal scale — one data dictionary instead of bespoke integrations.
- impact[1]
- Built the self-service portal that let trading-system teams request and manage their own middleware configuration, retiring a queue of manual tickets.
- impact[0]
- Shipped SIMN, a workflow engine for data-centre operators to author, validate, execute and audit network maintenance runbooks — 60%+ gain in operator efficiency.
- impact[1]
- Built a distributed reachability monitor with historical query APIs, letting alerting correlate application failures against device outages.
- impact[0]
- Reporting and anomaly-detection dashboards on Druid, plus an interface for managing Presto queries and automating Jenkins jobs for the data science and SRE teams.
- impact[0]
- Cut initial game load time 25% by preloading assets against player game state.
- impact[1]
- Dynamic task generation tuned to player skill: engagement +15%, retention +20% across demographics.
GET
/leadership
?level=INFO
I've moved from being the person expected to know how the system works to creating the conditions for a team to make good decisions, deliver reliably, and grow without depending on me for every answer.
DIRECTION
Direction beyond my own scope
When several vertical teams were each building their own SSE streaming APIs, I made the case for a shared Pub/Sub platform capability, built consensus across engineering managers, and turned it into a committed workstream — contracts agreed before implementation so consumers could validate the model early.
OWNERSHIP
Ownership placed, not accumulated
Offered entitlement, metering and throttling as one bundle of platform responsibility, I separated the problem and placed each capability with the organisation best positioned to own it. Scope discipline is a leadership decision, not a modesty one.
RISK
Risk named before it becomes failure
Used real delivery data to show a migration pace incompatible with the firm's timeline, surfaced capacity and key-person risk the backlog didn't represent, and converted it into a tiered headcount proposal with explicit trade-offs. Inside the team, KTLO, resiliency, modernisation and forward capacity are budgeted separately so priority is visible.
PROD
Ownership through production
Incidents driven to resolution and follow-up, structural issues tracked as risks with remediation plans, stakeholders told early enough to act. Formalised on-call and triage that had depended on senior engineers' institutional knowledge; the team's standard is detecting problems before clients do.
PEOPLE
Leadership capacity, built deliberately
Three senior engineers now independently lead substantial areas, represent the team cross-functionally and handle escalations; two carry broader Tech Rep responsibility, and one — after clearer expectations and deliberately chosen opportunities — is now technical lead and my backup. Specific feedback, ownership matched to development needs, visible contributions, then genuinely stepping back.
SEAMS
Work at the seams
Identity, configuration, permissioning and runtime infrastructure get hard at the boundaries between teams — that's where I've spent my time, including consolidating team-level risk submissions into organisation-wide themes so leadership could prioritise systemic issues instead of reconciling lists.
GET
/work
?selected=true
- platformEnterprise ingestion API platform
- observabilityIngestion visibility layer
- frameworkDataMapper
- self-serviceMiddleware portal
- automationSIMN workflow engine
Reveal impact ▾
Enterprise ingestion API platform
One platform for client data entering the firm — FIX, REST and BLPAPI on shared infrastructure, with pub/sub subscription management carrying 1M+ transactions daily at 60× the legacy throughput.
Ingestion visibility layer
Dashboards and interfaces that show clients and internal teams exactly how their data is flowing — latency patterns, error hotspots, per-client activity — turning support escalations into self-diagnosis.
DataMapper
A configuration-based data dictionary for transformation and translation across sources — 600+ requests per second, scaling horizontally instead of by hand-written adapters.
Middleware portal
Gave application teams direct control of their RabbitMQ and Kafka configuration, packaging and deployment — removing the middleware team as a bottleneck.
SIMN workflow engine
Authoring, validation, execution and audit for network device maintenance, turning tribal runbooks into repeatable workflows — 60%+ operator efficiency gain.
Key unlocks
1M+ txn / day
99.9% uptime · DR-1
60× throughput vs legacy
bi-weekly release cadence
10 team