● OPEN TO WORK · Jersey City, NJ · Software Engineer · Backend & Distributed Systems · AI
I build the backend systems and AI pipelines that turn messy data into fast, reliable software — Java/Python services, Kafka event platforms on AWS & Kubernetes, and production RAG with LLM agents. 5+ years across healthcare and finance.
This page isn’t a brochure — it runs scaled-down versions of systems I’ve actually built. Run a query, turn the dial, break something.
At McKesson I architected a GenAI document-intelligence platform (OpenAI + LangGraph agents + RAG) that extracts and validates unstructured medical records. Automating that path cut operational friction ~40% on the healthcare remediation platform.
At J.P. Morgan Chase I built the Datadog dashboards and Prometheus alerting across security microservices, and wired LLM-driven anomaly detection into compliance monitoring — halving the mean time to detect anomalies.
My cloud-native event-processing platform runs partitioned Kafka topics with Java consumer groups on Kubernetes and AWS Lambda for bursts. Idempotent consumers and disciplined offset management sustain 10k+ events/sec with zero message loss.
I hold a 90%+ JUnit/Mockito coverage baseline on the services I own, with CI quality gates — part of treating reliability and observability as the definition of done, not an afterthought.
/ 01 — RETRIEVAL
The RAG pipeline behind my GenAI document work and my Semantic Retrieval Engine. Pick a mode, run a query, and read the retrieval scores and per-stage latency.
/ 02 — THROUGHPUT
My Kafka event-processing platform — 10k+ events/sec, zero message loss. Drag the ingress load and watch consumers autoscale while p99 holds.
/ 03 — SELF-HEALING
My Self-Correcting Document Pipeline. Feed it a clean document, or a messy one and watch validation route it back for a self-correcting retry.
/ 04 — RESILIENCE
A request travelling through a distributed backend I’d build. Send it, then inject a failure and watch retries, replica promotion and caching keep it alive.
/ 05 — TERMINAL
A real command parser. Try skills · projects · benchmark · kubectl get pods · deploy. help lists everything.
/ 06 — BUILT
Two open-source builds you can read on GitHub, plus production and independent systems. Click any card to expand the full case study.
RAG retrieval shipped as reusable MCP infrastructure — index a corpus once, query it from any AI client.
Hybrid vector + BM25 fused with Reciprocal Rank Fusion; interchangeable SQLite / Postgres+pgvector stores; pluggable OpenAI or keyless embeddings; a recall@k eval harness gates CI.
Multi-agent document processing on LangGraph that repairs its own failures.
classify → extract → validate → summarize with a confidence-weighted retry loop; low-confidence docs route to review instead of fabricating fields; every node is auditable.
LLM agents + RAG turning unstructured medical records into structured data.
OpenAI models and LangGraph agents over a RAG pipeline; parse → chunk → embed → semantic search → extraction; delivered via REST/gRPC on Kubernetes, monitored with Datadog + Prometheus.
LLM anomaly detection inside security compliance monitoring.
High-frequency Kafka audit streams into concurrent Java parsers; an LLM layer flags anomalous access patterns; low-latency gRPC services on tuned Oracle/SQL Server with Datadog alerting.
Semantic search + extraction with RAG and agent orchestration.
Java (Spring Boot) owns a secure REST/gRPC surface with strict RBAC; a Python (FastAPI) service runs LangGraph agents over RAG; embeddings in Pinecone + pgvector; full-path observability.
Event-driven microservices at 10k+ events/sec with zero message loss.
Partitioned Kafka topics; Java consumer groups on Kubernetes; AWS Lambda absorbs bursts; DynamoDB state; idempotent consumers and offset discipline give at-least-once with dedup.
/ 07 — TRACK RECORD
Three companies, one throughline: production backend and AI at healthcare and bank scale. The numbers below are what I shipped — not what I aspire to.
/ 08 — BLUEPRINT
A representative production shape from systems I’ve built — not any employer’s proprietary diagram.
/ 09 — SIGNAL
Open to Software / Backend / Senior SWE and AI / Applied AI / Generative AI Engineer roles.