Last login: Mon Apr 14 09:32:01 on ttys001
pradeep@portfolio:~$
> Pradeep Prakash
Backend & AI Platform Engineer · LLM Observability · Bengaluru / Remote · 7+ years
open to backend, platform, and AI infrastructure roles
~
now:
building LLM observability and migration reliability systems
pradeep@portfolio:~$
Building reliable backend and AI platform systems with observability, data pipelines, and developer workflow guardrails.
browse my work, download my resume, or ask my AI agent about migrations, reliability, and LLM systems.
pradeep@portfolio:~$
▋
pradeep@portfolio:~$
{
"name": "Pradeep Prakash",
"role": "Backend & AI Platform Engineer",
"company": "Fynd (Reliance Retail)",
"location": "Bengaluru, India",
"remote": "open_to_remote",
"experience": "7+ years",
"focus": [
"llm and agent observability",
"fault-tolerant data and migration systems",
"ai-assisted engineering workflows",
"end-to-end reliability ownership"
],
"status": "open_to_opportunities"
}
pradeep@portfolio:~$
Backend and platform engineer with 7+ years building high-throughput data and migration systems, with recent focus on LLM/agent observability and AI-assisted developer workflows. Reliability-oriented and comfortable owning systems end-to-end across design, implementation, on-call, and RCA. Strong in Node.js, TypeScript, React, Kafka, MongoDB, PostgreSQL, Redis, Kubernetes, and GCP/AWS.
pradeep@portfolio:~$
Node.js
TypeScript
React
MongoDB
Redis
Kafka
Kubernetes
GCP
AWS
→Architected AgentSentinel with a trace/span model for agent runs plus LLM/tool/DB spans; backed by 138 frontend and 416 backend tests
→Built a migration validation workbench with deterministic checks, an optional LLM audit layer, and 52 regression tests across 9 files
→Shaped an AI-assisted engineering operating model with path-scoped context, review and debug agents, syntax hooks, project memory, and deploy safeguards
→Built a 6-stage ingestion pipeline processing 300 records/sec at p99 under 400 ms with checkpointing, DLQs, and idempotent writes across 1M+ records
→Cut peak pod memory 80% (1.5 GB → 300 MB) on 200K-row jobs via cursor-based streaming and adaptive chunk sizing
→Reduced production incidents by 40% via circuit breakers, health-signal probes, and bounded error propagation
→Owned Fynd Coupons from 0 to 97K MAU with APIs, Redis-backed caching, rate limits, 300 req/s at p99 under 300 ms, and 99.9% availability
→Tech-led Fynd Migrate across a small team serving 50+ merchants, replacing a 2–3 day manual onboarding flow with self-serve migration
Node.js
PostgreSQL
Redis
React
Microservices
→Designed and owned Teacher Tech for 10,000+ tutors; reduced payroll from 3–4 days to 1 day and removed the engineering bottleneck for 3 ops teams
→Built Tutor CMS in React, moving routine academic-content updates from dozens of engineering tickets per month to near-zero manual dependency
Software Engineer @ Acube Tech May 2020 – Jun 2021
Built backend for a healthcare teleconsultation platform supporting concurrent real-time patient-to-expert sessions with high availability.
Product Engineer @ Codingmart Oct 2018 – May 2020
Built event-driven microservices on RabbitMQ for BookMyShow's ticketing stack, decoupling booking, inventory, and notification flows for fault isolation and throughput.
pradeep@portfolio:~$
Internal LLM and AI-agent observability plus cost-intelligence platform with a trace/span model for agent runs, LLM spans, tool spans, and database spans.
138 FE tests
416 BE tests
trace/span model
TypeScript
React
LLM
Observability
Deterministic field-level checks as the trust anchor, with an optional LLM audit layer for deeper review across migration validation flows.
6 entity types
52 tests
trust-first checks
Node.js
TypeScript
Validation
LLM Audit
Shopify → Fynd migration platform serving 50+ merchants with 1M+ records migrated and self-serve onboarding replacing a manual migration flow.
300 rec/s
p99 < 400ms
1M+ records
Node.js
Kafka
MongoDB
Kubernetes
Onboarding, scheduling, audit, and payroll platform for 10,000+ tutors, reducing payroll processing to 1 day while removing operational bottlenecks.
10K+ tutors
1-day payroll
3 ops teams
Node.js
PostgreSQL
React
pradeep@portfolio:~$
▸ AI / LLM SYSTEMS
LLM Observability
Agent Tracing
Token & Cost Analysis
AI-assisted Dev Workflows
Deterministic + LLM Validation
▸ BACKEND & APIs
Node.js
TypeScript
REST APIs
GraphQL
Microservices
▸ STREAMING & MESSAGING
Kafka
RabbitMQ
Event-driven Systems
▸ DATA & INFRASTRUCTURE
MongoDB
PostgreSQL
Redis
Docker
Kubernetes
GCP
AWS
CI/CD
▸ FRONTEND & INTERNAL TOOLS
React
Dashboards
State Management
Performance Optimization
▸ RELIABILITY
On-call
RCA
Circuit Breakers
Dead-letter Queues
Idempotency
Back-pressure