Satyakam Singhal

India's largest debt marketplace for startups (₹3,000Cr+ deployed) — core engineer on AICA, the AI credit-assessment platform

• Built a template-driven document-generation platform (template registry, versioned variable catalog across company/financial/GST data, audit trail, PDF/DOCX export) — cut new document-type rollout from 2–3 weeks of engineering to same-day configuration.
• Led migration of the platform's rich-text editing core across every document surface after auditing 19 recurring defect classes — server-side PDF export, DOCX import, AI-assisted drafting, and a bounded autosave pipeline; eliminated the platform's largest source of user-reported bugs.
• Cut underwriting data costs by making compliance screening (legal/AML/bureau) configurable per partner (~20–30% fewer third-party API calls) and moving bank-statement ingestion on-demand — ending vendor spend on leads that never convert.
• Revamped the lender-facing due-diligence report: 20 sections restructured into lender narrative order, per-partner configurability without code releases, 30% faster load.
• Resolved 90+ production incidents across Java/Spring Boot, Node, and Python microservices; built multi-agent LLM pipelines for SDLC and QA (architecture→implementation→review orchestration, API test automation) used by the team daily.

Third hire at a consumer-lending fintech; owned the platform end-to-end through to acquisition by Crib (proptech)

• Architected the lending platform from scratch — onboarding journey, database design, KYC/AML integrations, CI/CD, and business-rule engines for real-time credit decisioning on 100+ data points.
• Scaled to ₹20Cr+ lending volume and 1,000+ active loans; led an 8-person team and onboarded 20+ B2B partners through a self-serve SaaS dashboard and plug-and-play lending SDK.
• Cut third-party API costs 50% with a data-standardization, caching, and fallback layer (Redis, S3, DynamoDB) — plug-and-play vendor onboarding and stronger fault tolerance.
• Grew Google impressions 100K → 6M+/month and organic traffic 0 → 10K/month by engineering a programmatic-SEO pipeline over real-time property data — brand growth at zero acquisition cost.
• Stood up platform-wide observability (PostHog, SigNoz): real-time alerts, crash monitoring, and performance tracing for proactive incident resolution.

• Cut bulk OCR document processing from 12 min to under 2 min per 30-page PDF by parallelizing OCR tasks and caching redundant executions on AWS Lambda.
• Replaced a paid Azure API with an in-house multilingual NLP transliteration engine — recurring cost to zero, several-fold faster, no accuracy tradeoff.