Headline

Senior Software Engineer | Python, React, Distributed Systems, System Design | Data pipelines, ClickHouse, LLM/RAG | 7+ yrs building reliable data-intensive systems

About

I'm a backend engineer with 6+ years building data-intensive systems and the pipelines and APIs that keep them reliable at scale. Today I own core data infrastructure at Cloudverse, a multi-cloud FinOps platform. I built the pipeline that ingests AWS Cost & Usage Report data from customer S3 buckets into ClickHouse — orchestrated with Airflow, autoscaled on Kubernetes with KEDA, and designed for exactly-once processing so nothing is double-counted or dropped. I also built the Reserved Instance / Savings Plan recommendation engines that turn raw billing data into concrete savings for customers across AWS, Azure, and GCP, and tuned ClickHouse to serve analytics dashboards in under three seconds. What I care about most is the unglamorous stuff that decides whether a system survives production: idempotency, retries, clean data models, and predictable behavior under load. On the side, I built an end-to-end, LLM-powered legal research platform that makes millions of Indian High Court judgments searchable — a schema-constrained extraction pipeline over locally-served models, semantic search with pgvector, and RAG chat with grounded, non-hallucinating answers, all on local hardware with no cloud LLM dependency. I'm strongest in Python (including the internals — asyncio, the GIL, generators, context managers), distributed job orchestration, and reliability-focused system design. 📫 avinash.kumar2020@outlook.com · GitHub: github.com/a-v-i-7

Location

Greater Bengaluru Area

AI stack

antigravity

claude code

codex

deepmind

deepseek

gemini

Work experience

Senior Software Engineer at TsquaredC, Apr 2026 - Present · 7 mos

Full-stack Developer at TsquaredC, Apr 2025 - Mar 2026 · 1 yr, -Built a scalable ingestion pipeline that pulls AWS CUR data from customer S3 buckets into ClickHouse — orchestrated with Airflow, autoscaled on Kubernetes via KEDA on RabbitMQ queue depth, with file-level parallelism for large multi-tenant datasets. -Designed the reliability layer for exactly-once processing: a Postgres processed_files table enforcing idempotency and dedup on (org_id, s3_etag), with RabbitMQ acked on receipt and Postgres owning the full job lifecycle including retries. -Developed multi-cloud RI / Savings Plan recommendation engines: extract uncovered on-demand spend from normalized CUR, simulate hourly commitment levels against live pricing APIs, and generate Compute/Database SP and Databricks utilization recommendations. -Engineered AWS CUR normalization (allocation pools + two-layer pricing-model resolution) as the foundation for cost allocation. -Optimized ClickHouse dashboards to sub-3s loads (SummingMergeTree materialized views, projections, month partitioning, LowCardinality, tiered query architecture + Redis caching).

Co-Founder at BillionScale, Sep 2020 - Dec 2024 · 4 yrs 4 mos, Bengaluru, Karnataka, India, -Co-founded and led engineering for an early-stage startup shipping multiple production apps, owning architecture, infrastructure, and deployment end to end. -Designed and shipped a digital canteen-management system (React, Django, GraphQL, PostgreSQL) serving daily student and staff transactions. -Optimized API performance and caching to cut latency across services; led deployment on DigitalOcean via Docker and Nginx.

Frontend Developer at actyv.ai®, Feb 2020 - Aug 2020 · 7 mos, Bengaluru, Karnataka, India, -Designed and shipped a DigiLocker API microservice (Node.js, Redis, MongoDB) integrating with India's national digital-identity platform, which serves hundreds of millions of users. -Integrated additional third-party APIs, wrote technical documentation and user stories, and worked in an Agile Scrum team.

Education

Bachelor of Technology - BTech, Computer Science, CMR Institute Of Technology