Adhiraj Bhattacharya profile photo 1 on Tal

Adhiraj Bhattacharya

Lead Software Engineer @ Zeta | System architect & hands-on technical lead, financial-grade distributed systems | Architected an EOD platform proven to close 208M+ accounts in under an hour | ex-Citi | NIT Durgapur

Bengaluru

13+ years of work experience

Zeta
Jan 20224 yrs 9 mosPresent

Bengaluru, Karnataka, India

System architect and hands-on technical lead — EOD accounting platform for credit instruments, in production at HDFC Bank. Technical direction of a ~15-engineer platform group. • Architected the EOD platform proven to close 208M+ accounts in under an hour at 1M / 100K peak TPS (processing / persistence) in load tests and customer demos — ~100× up from a ~2M ceiling, by decoupling processing from persistence. • Personally built the big-data data-prep pipeline (S3 + Athena + Apache Iceberg): 30 min / 100K accounts → <5 min / 200M+; nightly-close infra cost −70–85% vs a scale-up Postgres replica; pre-run prep removed live network-fault errors from the close. • Designed a reusable batch command framework (workers as pure functions emitting commands, persisted only on full-job success) — atomic close, independently scalable persistence; adopted org-wide. • Collapsed 18 per-worker deployments across 6 phases into one deployment with a linear per-phase runner. • Replaced a serverless real-time event processor (cold starts, slow scale-up) with an always-on, horizontally scaling Kafka-fed REST service plus cron retry over persisted failures — transient failures self-heal; only unhandleable ones reach a human. • Turned a per-call REST classification lookup into an in-process cached library (network hops to zero; adopted by every classifying system); cut account-creation setup latency ~4 s → <200 ms (~20×). • Built an extensible test-environment automation framework (bring-up → seed → execute → validate → tear-down): ~3-day manual validation → ~4 h (−95%). • Initiated and led org-wide adoption of agentic engineering tooling — conceived the first three skills: the de-facto SRE release pre-check, a code-review gate for AI-generated code, and Snyk auto-remediation. • Set org-wide standards: CI quality/security gates on every PR (Jenkins, SonarQube, Snyk); lean production logging. • Mentored 10+ engineers to senior.
Citi
Jun 20183 yrs 7 mosDec 2021

Pune, Maharashtra, India

Technical lead — regulatory balance-sheet forecasting engine (US CCAR). • Made the forecasting engine's reference rate configurable — added transaction-based alternative rates alongside LIBOR/SOFR so the engine moves off LIBOR by configuration, not code. • Made runtime-tunable configuration live — changes apply at runtime with no application restart. • Decoupled input-data validation into a standalone precheck service, taking validation runtime off the forecasting run's critical path; runs proceed only on validated data. • Automated Long-Term-Debt risk calculation by integrating a third-party analytics engine in place of a manual Excel process (zero manual intervention), verified against seeded expected results with a repeatable regression suite. • Took the codebase from 0 to ~80% test coverage — set the standard, led the team to it, and enforced it with a JaCoCo coverage gate in the build.
Principal Financial Group
Mar 20144 yrs 3 mosMay 2018

Pune Division, Maharashtra, India

End-to-end marketer services platform: onboarding, commission (variable-payout) calculation, statement generation, disbursement. De-facto tech lead of a ~6-person team — owned technical design, architecture and task breakdown. • Modernised the commission-calculation, statement and disbursement flows — migrated legacy COBOL/DB2 into a modular Java application, unifying separate flows and reducing statement/processing runtime (statement prep moved to MongoDB with a batched templating/dispatch component). • Redesigned marketer onboarding — a 20-form flow cut to 4 through reuse and redesign, cutting onboarding lead time.
Mjunction Services ltd (SAIL-TATA Steel venture)
Jul 20138 mosFeb 2014

Kolkata, West Bengal, India

Proprietary auction engine for SAIL to trade steel byproducts.

Adhiraj’s AI Stack

claude code
claude code
claude
claude

Education

National Institute of Technology Durgapur

Bachelor of Technology (BTech), Electrical Engg.

National Institute of Technology Durgapur

20094 yrs2013

Join Adhiraj and 10k+ other professionals on

Tal Boss

Tal is the app where bosses meet engineers & builders directly, without middlemen

Download Tal Boss

Adhiraj Bhattacharya on Tal

Headline

Lead Software Engineer @ Zeta | System architect & hands-on technical lead, financial-grade distributed systems | Architected an EOD platform proven to close 208M+ accounts in under an hour | ex-Citi | NIT Durgapur

About

Backend architect and hands-on technical lead with 13+ years building financial-grade, high-throughput distributed systems. At Zeta I architected an end-of-day (EOD) accounting platform for credit instruments, proven in load tests and customer demos to close 208M+ accounts in under an hour at 1M / 100K peak TPS (processing / persistence) — a ~100× scale-up from a ~2M-account ceiling, achieved by decoupling processing from persistence so each scales independently. I own system design end-to-end and stay in the code: I design the architecture and build the hardest parts myself (for example the S3 + Athena + Apache Iceberg data-prep pipeline that cut nightly-close infra cost by ~70–85%), while setting engineering standards and mentoring a ~15-engineer platform group to adopt them (10+ engineers mentored to senior). Delivered to HDFC, ICICI, Axis (India) and PNC Bank (USA). Most recently I initiated and led org-wide adoption of agentic engineering tooling — conceiving and building the first three "agent-harness skills" myself, including the deployment pre-check the SRE team now runs before every release. Before Zeta: regulatory balance-sheet forecasting (CCAR) at Citi — configurable reference rates for the move off LIBOR, live runtime configuration, and a codebase taken from 0 to ~80% test coverage; and at Principal, modernising a COBOL/DB2 commission-and-payout platform into a modular Java application. Core stack: Java, Spring, SQL · PostgreSQL, MongoDB, Redis · Kafka, Apache Flink, Apache Iceberg, S3, Athena · AWS, Docker, Kubernetes · CI quality/security gates (Jenkins, SonarQube, Snyk), test automation, agentic/LLM tooling, cost optimisation. Open to Technical Lead, Staff/Principal Engineer and Architect roles on distributed systems and platforms — and to Engineering Manager roles where hands-on architecture still matters.

Location

Bengaluru

AI stack

claude code

claude

Work experience

Lead Software Engineer (System Architect / Technical Lead) at Zeta, Jan 2022 - Present · 4 yrs 9 mos, Bengaluru, Karnataka, India, System architect and hands-on technical lead — EOD accounting platform for credit instruments, in production at HDFC Bank. Technical direction of a ~15-engineer platform group. • Architected the EOD platform proven to close 208M+ accounts in under an hour at 1M / 100K peak TPS (processing / persistence) in load tests and customer demos — ~100× up from a ~2M ceiling, by decoupling processing from persistence. • Personally built the big-data data-prep pipeline (S3 + Athena + Apache Iceberg): 30 min / 100K accounts → <5 min / 200M+; nightly-close infra cost −70–85% vs a scale-up Postgres replica; pre-run prep removed live network-fault errors from the close. • Designed a reusable batch command framework (workers as pure functions emitting commands, persisted only on full-job success) — atomic close, independently scalable persistence; adopted org-wide. • Collapsed 18 per-worker deployments across 6 phases into one deployment with a linear per-phase runner. • Replaced a serverless real-time event processor (cold starts, slow scale-up) with an always-on, horizontally scaling Kafka-fed REST service plus cron retry over persisted failures — transient failures self-heal; only unhandleable ones reach a human. • Turned a per-call REST classification lookup into an in-process cached library (network hops to zero; adopted by every classifying system); cut account-creation setup latency ~4 s → <200 ms (~20×). • Built an extensible test-environment automation framework (bring-up → seed → execute → validate → tear-down): ~3-day manual validation → ~4 h (−95%). • Initiated and led org-wide adoption of agentic engineering tooling — conceived the first three skills: the de-facto SRE release pre-check, a code-review gate for AI-generated code, and Snyk auto-remediation. • Set org-wide standards: CI quality/security gates on every PR (Jenkins, SonarQube, Snyk); lean production logging. • Mentored 10+ engineers to senior.

Lead Software Engineer at Citi, Jun 2018 - Dec 2021 · 3 yrs 7 mos, Pune, Maharashtra, India, Technical lead — regulatory balance-sheet forecasting engine (US CCAR). • Made the forecasting engine's reference rate configurable — added transaction-based alternative rates alongside LIBOR/SOFR so the engine moves off LIBOR by configuration, not code. • Made runtime-tunable configuration live — changes apply at runtime with no application restart. • Decoupled input-data validation into a standalone precheck service, taking validation runtime off the forecasting run's critical path; runs proceed only on validated data. • Automated Long-Term-Debt risk calculation by integrating a third-party analytics engine in place of a manual Excel process (zero manual intervention), verified against seeded expected results with a repeatable regression suite. • Took the codebase from 0 to ~80% test coverage — set the standard, led the team to it, and enforced it with a JaCoCo coverage gate in the build.

Senior Analyst (Senior Software Engineer → Technical Lead) at Principal Financial Group, Mar 2014 - May 2018 · 4 yrs 3 mos, Pune Division, Maharashtra, India, End-to-end marketer services platform: onboarding, commission (variable-payout) calculation, statement generation, disbursement. De-facto tech lead of a ~6-person team — owned technical design, architecture and task breakdown. • Modernised the commission-calculation, statement and disbursement flows — migrated legacy COBOL/DB2 into a modular Java application, unifying separate flows and reducing statement/processing runtime (statement prep moved to MongoDB with a batched templating/dispatch component). • Redesigned marketer onboarding — a 20-form flow cut to 4 through reuse and redesign, cutting onboarding lead time.

Software Engineer at Mjunction Services ltd (SAIL-TATA Steel venture), Jul 2013 - Feb 2014 · 8 mos, Kolkata, West Bengal, India, Proprietary auction engine for SAIL to trade steel byproducts.

Education

Bachelor of Technology (BTech), Electrical Engg., National Institute of Technology Durgapur

Adhiraj Bhattacharya - Lead Software Engineer (System Architect / Technical Lead) in Bengaluru | Tal