Headline

Builder | Ex-Goldman Sachs, Google, slice | IITR | CFA L2

About

I build systems where being wrong is expensive — money moving between banks, revenue books auditors sign off on, LLM output that has to be right, not just plausible. I get pulled into ambiguous domains. Card management systems had no public playbook, I reconstructed one through primary research — it became the backbone of slice's card program. At Google, I built GCP's revenue-recognition platform — turning accounting policy into APIs, data models, and a fully automated system. The recent chapters are LLM-native: building agent workflows that decompose real analyst and QA work into systems with measurable, guarded output. The breadth is deliberate: markets at Goldman Sachs, cloud platforms at Google, payments at slice, AI at Peekstreet — which I co-founded — and Testsigma. Founding a company taught me to own outcomes end to end, and I've kept the habit. I love owning hard problems.

Location

Bengaluru

AI stack

claude code

claude

codex

cursor

hermes agent

Work experience

AI Lead at Testsigma, Sep 2025 - Present · 1 yr 1 mo, ◦ Build a multi-stage LLM agentic pipeline that generates test cases from product requirements, with built-in LLM evaluation (evals), guardrails, and context engineering, cutting test authoring time by 75%. ◦ Directed a knowledge-graph prototype that grounds test generation in each customer’s application structure for hallucination mitigation. ◦ Designed an auto-healing algorithm to repair broken test-automation locators across heterogeneous customer applications. ◦ Mentored engineers implementing backup locators and multimodal LLM-based visual analysis with confidence scoring. ◦ Reduced false test failures by 82% across all customers while keeping false positives under 5%. ◦ Provided technical leadership to 10+ engineers across multiple cross-functional teams, shipping 5+ projects through influence without authority.

Founder at Stealth Startup, Apr 2024 - Oct 2025 · 1 yr 7 mos, Bengaluru, • Built and launched a multi-agent generative AI platform (autonomous LLM agents) for investment research and analysis. • Designed an agent-orchestration framework from scratch (zero-to-one), decomposing complex analysis workflows into specialized sub-agents with task-specific tools (tool use / function calling). • Developed a personalization layer that adapts agent output (structured outputs) to each client’s investment mandate and reporting format. • Implemented a Retrieval-Augmented Generation (RAG) system — document parsing, chunking, embeddings, and vector search — for fast, accurate retrieval over earnings call transcripts. • Led customer-facing design-partner collaboration to automate earnings transcript analysis, cutting analyst effort by 33%.

Technical Lead at slice, May 2022 - Apr 2024 · 2 yrs, Bengaluru, - Card Management System (CMS), PPI card ◦ Architectured CMS through primary research and vendor evaluation with RuPay, VISA, vendors, and competitors. ◦ Synthesized the research into a buildable high-level architecture and technical specification that became the blueprint for the entire CMS program. ◦ Scoped all components: multi-tenant management, digital and physical card issuance, card controls and transaction limits, online and offline transaction authorization. ◦ Specified entities, domain models, state machines, data flows, card lifecycle management, and algorithms for every user journey, turning an ambiguous domain into an executable design. ◦ Deployed HSMs for all cryptographic operations (encryption, key management) across DC/DR data centers with disaster recovery, meeting PCI-DSS and RBI regulatory compliance requirements. ◦ Designed a modular monolith with clean service boundaries (domain-driven design), enabling incremental decomposition into microservices for scalability. ◦ Future-proofed the design for extensibility to credit and debit cards and additional card networks (NPCI/RuPay, Visa), avoiding re-architecture for new products. ◦ Projected cost savings of INR 30M annually at launch volume, scaling to INR 100M+ as card volume grows. - Payouts ◦ Built the loan disbursement (lending payouts) platform with bank integrations, defining bank account discovery, penny-drop validation, and transaction behavior. ◦ Designed client-level idempotency and pending-transaction balance locking (concurrency control, transactional integrity), eliminating negative-balance and double-disbursement scenarios. ◦ Implemented event-driven transaction state management using webhooks and asynchronous processing, with end-of-day monetary reconciliation against the ledger.

Software Engineer at Google, Oct 2020 - Apr 2022 · 1 yr 7 mos, Bengaluru, ◦ Built GCP’s revenue recognition system, achieving 100% compliance automation. ◦ Reduced manual intervention by 70% by partnering cross-functionally with product managers and finance leads to translate business requirements into technical solutions. ◦ Delivered 20+ REST APIs with end-to-end architecture, including entities, data modeling, low-level design, API contracts, and business logic. ◦ Designed and built a changelog-capture (audit trail) system, engineered for auditability, extensibility to future business needs, and scale.

Senior Software Engineer, Global Investment Research at Goldman Sachs, Jun 2018 - Sep 2020 · 2 yrs 4 mos, Bengaluru, - Quant ◦ Built statistical and time-series quantitative models computing the Financial Conditions Index (FCI) for 50 countries. ◦ Validated the models via backtesting against 20+ years of history, demonstrating correlation with real economic indicators. ◦ Engineered a real-time data pipeline (MongoDB) delivering FCI data directly to research analysts. ◦ Shipped production REST APIs powering the public FCI dashboard on goldmansachs.com and external client data feeds. - Alternative data modelling ◦ Estimated country-level import/export trends by tracking export-vessel locations (alternative data, predictive modeling). ◦ Modeled the impact of corporate events on stock prices (financial modeling). ◦ Developed Natural Language Processing (NLP) text analytics for trend analysis over historical earnings transcripts, adopted by analysts across the division. ◦ Trained 100+ fundamental analysts to build their own Python data-analysis workflows, seeding automation across the research division.

Software Engineer, Market Risk at Goldman Sachs, Jun 2017 - May 2018 · 1 yr, Bengaluru, • Built data analytics tooling over daily risk-pricing data for the firmwide portfolio. • Reduced the compute cost and runtime of risk-pricing calculations (performance and cost optimization).

Education

Bachelor's Degree, Computer Science, Indian Institute of Technology, Roorkee

High School, Step By Step High School, Jaipur