01● LIVE
deadlock-dinner
🍝 Interactive, kid-friendly visualization of the Dining Philosophers problem — locks, mutexes, semaphores & deadlock come alive, with real Java & Go code.
JavaScript · concurrency viz
01 identity record
A poet who ships RAG.
I hold the architecture and the taste; the machine holds the boilerplate. Production GenAI on distributed infrastructure — and a soft spot for things that bend through the air.
| handle | @chitransh-saxena |
|---|---|
| role | Sr. Staff SWE — IBM |
| focus | RAG · embeddings · streaming |
| base | Bangalore, IN · UTC+5:30 |
| exp | 7+ yrs · 3+ GenAI · 1.5+ Claude Code |
| tooling | Claude Code daily, 1.5+ yrs |
| status | ◉ open to hard problems |
- signal RAG · +35% recall@20
- scale 1.5M txns/day · 99.99%
- off-hours FPV drones · guitar · photography
- on grass trivela free-kicks · juggling
02 experience.ledger — open a row
Three latent shelves.
01
IBM Software Labs
Sr. Staff SWE
2024 — now
BLR
RAG · +35% recall@20
watsonx Code Assistant for Z — configurable SaaS embedding & RAG platform · Go codebase
- Architected & led (4 devs · 2 QEs) an end-to-end embedding + RAG pipeline ingesting 8K+ code pairs, 50+ docs (500+ pages each) and 250+ runtime examples.
- Integrated SPLADE learned-sparse retrieval after a deep BM25 study → +35% recall@20 with sub-second hybrid inference.
- Prototyped vector search on MilvusDB (HNSW + quantization, IVF, <900ms inference), then ported to a hybrid Elasticsearch on x86 + OpenSearch on s390x (IBM Z) stack — dual-architecture, dual-store.
- Redis Streams backbone at sub-100ms / 99.95%; novel LSP-based chunking over 10K+ artifacts via custom MCP tools.
- Re-architected a monolith → 90% reuse, 100% throughput; adopted by 4+ teams. ★ Star of the Month in 2 months.
- Synthetic-data pipeline: 50K pairs from 5K (10×), TDD-validated, for LLM extended pre-training.
- Shipping it in Go (Golang) — actively leveling up the language on the job.
GoRAGSPLADEMilvusDBElasticsearchOpenSearchs390xRedis StreamsMCP
02
Walmart Global Tech
SWE 3
2021 — 2024
BLR
1.5M txns/day · 99.99%
Fault-tolerant event-driven stream processing at retail scale
- Stream platform sustaining 1.5M txns/day @ 50 TPS on 3-broker Kafka (RF=3) — exactly-once, DLQ isolation, 99.99% SLO.
- Cut p99 consumer lag 2s → 0.5s; collapsed 15+ queries into 3; wire format −67% (150 → 50 bytes).
- Plugin framework for 50+ event types / 15+ topics (Strategy/Factory/Decorator + Resilience4j).
- 12+ autoscaled microservices on K8s; MERN control plane cut releases 7–10 days → 45s across 20+ repos.
- Led SRE P1 response → MTTR −40% (centralized logging + Jaeger tracing).
KafkaExactly-onceKubernetesResilience4jJaegerMERN
03
NielsenIQ
SWE
2019 — 2021
MAA
C++ −55% runtime
Decomposing monoliths, squeezing C++ pipelines
- Decomposed 3 monoliths → distributed microservices w/ leader election → throughput +60% (100 → 160 RPS).
- Re-engineered latency-critical C++ pipelines −55% (20s → 9s) across 15K+ LOC — profiling, leak elimination, hot-path restructuring; owned on-call.
C++MicroservicesLeader electionOn-call
edu — B.Tech Information Technology · SRM Institute · 2015–2019 · 88%
03 stack.matrix — hover to ignite a cluster
The competency lattice.
00GenAI / MLRAGEmbedding modelsSemantic & hybrid searchBM25 / SPLADERe-rankingMultimodal ingestSynthetic dataFine-tuningPrompt engAgentic codingMCP11
01Retrieval / ANNFaissHNSWIVFQuantizationANN indexingHybrid retrieval06
02DatabasesPostgreSQLMS SQL ServerMongoDBElasticsearchOpenSearchMilvusDB06
03Streaming / DistributedApache KafkaRedis StreamsExactly-onceEvent-drivenDLQsMicroservicesLeader electionCircuit breakersHASRE & observability10
04FrameworksSpring BootSpringResilience4jgRPCREST APIsNode / ExpressReact07
05LanguagesJavaPythonC++GoSQLNoSQL06
06InfrastructureKubernetesDockerHelmAzureGrafanaPrometheusJaeger07
07Personal / Self-hostedGitHub SCMGitHub ActionsGitHub IssuesGitHub ProjectsCloudflare DNSCloudflare WorkersCloudflare Zero TrustWAF / SSLOracle Cloud · OCIDockerTerraformWrangler12
04 how i build
I ship with Claude Code in the loop.
For 1.5+ years it's been my default tooling — agentic coding, custom MCP servers, rapid POCs across enterprise & personal work. I hold the architecture and the taste; it holds the boilerplate. More shots on goal.
- POCs in hours, not weeks
- Custom MCP tools wired into real data
- Tests-first, even for throwaways
~/portfolio claude "render my résumé as a latent space" ✓ 24,000-point embedding cloud · one draw call ✓ cursor → query vector · kNN retrieval in-shader ✓ Instrument Serif · Hanken Grotesk · IBM Plex Mono → shipped to the edge ~/portfolio ▍
05 projects.index · 2 live, more incoming
Sub-constellations.
02● LIVE
trivela
⚽ A free-kick game that makes the Magnus effect impossible to misunderstand — curl it round the wall into the top bins. Vanilla web component, embeddable; play it live or read the source.
JavaScript · canvas · physics
03PENDING
project-three
something with range — a distributed toy / viz / tool. Indexing now.
tbd · tbd
06 arcade — a tiny game, built into the page
Don't clip the hurdles.
07 library — recent reads · pick a book
The shelf lattice.
08 open channel
Got a hard problem in
GenAI or distributed systems?
| → | GitHub | github.com/Chitransh-Saxena ↗ |
| → | linkedin.com/in/chitransh22 ↗ | |
| → | LeetCode | leetcode.com/u/chitransh22 ↗ |
| ▦ | Résumé | preview · download pdf ↓ |