Blog

A running log of what I’m building, learning, and occasionally over-engineering. If you’re here : hi! this is the human side of the resume.

Work & what I’m up to

Notes from the trenches : microservices, observability, and a soft spot for clean APIs.

Oracle ~ Associate Consultant

Jul 2023 – Present • Pune, India

Day-to-day, I live in Java/Kotlin microservices on OCI. Think REST + gRPC, a lot of tracing, and the occasional “why is this one pod special?” mystery. I like turning paths into boring ones, the sort that never page you at 3am.

On the Batelco side (Jan ’25 →), I’ve been shaping 30+ services, tightening interfaces into the 5G core (NRF, PCF) and Wi-Fi control plane, and rolling them out with Helm + Istio. Prometheus says we’re at +37% throughput and −12% latency. I’ll take it.

Earlier (PPC Retail), I worked full-stack with OJET + Node.js, wrote reusable BRM opcodes, stood up SSR/serverless flows with 30+ APIs, and shaved load everywhere I could. I also documented obsessively, enough that new folks saved ~160 hours getting onboarded.

Java • Kotlin • gRPCOCI • Helm • IstioPrometheus/GrafanaOJET • Node.js

Research internship (Nibha Tech)

Dec 2022 – May 2023 • Nashik, India

I helped build a national mutual funds performance tool; biggish time-series (25 years), sensible features (>30k vectors), and models that behave under stress. The goal was practicality: explainable signals (SHAP), guarded optimism (GAN augmentation), and benchmarks we could defend (MLflow).

TL;DR: 96.2% in normal regimes, 76.03% in simulated crises, plus real-time APIs the advisors actually used.

Interests (tech & otherwise)

Stuff I keep circling back to.

Systems & reliability: clean interfaces, noisy graphs, and the satisfaction of turning a flaky path into a boring one. Observability is a feature.

ML that ships: agents that read docs so humans don’t have to; models that explain themselves; sensible guardrails. (Yes, I like LangChain when it’s scoped well.)

Edge & embedded: low-power tricks, RL on tiny footprints, energy harvesting; making small devices surprisingly opinionated.

Non-tech: chess, swimming, cricket, pickup basketball; planting trees with community folks; the kinds of weekends that smell like soil and sunscreen.

API designPerf tuningGenAI agentsTime-seriesEmbedded RL

Certificates

Full list lives on LinkedIn; raw files live in Drive. Highlights below link to the same.

Community & volunteering

GDSC (VIT Pune), Team Head: ~130 hours of sessions on ML, Flutter, Web3; a lot of Q&A, a lot of “it depends.”

Ekasutram: wrote 100+ problems, ran ML workshops, and generally encouraged people to try one more idea than they planned to.

Mentorship: helped 70+ freshmen find their footing, the best metric is when you’re no longer needed.

Team Green / NSS / NGO “We Listen” / NCC: trees, teaching, and a bit of drill; the discipline sticks, and so do the habits.

Now (Aug 2025)

Tuning microservices around 5G control-plane edges, experimenting with leaner doc-reading agents for onboarding, and writing fewer but clearer internal notes.

If you’re reading this and want a deeper dive, I’m happy to share design docs or experiment logs, the interesting parts are always in the trade-offs.