Headline

Backend Software Engineer (SDE-1) @ HealthifyMe | Ex-Amazon SDE | Python · Django · FastAPI · AWS · Kafka | Production LLM systems: grounded AI diet plans ~900/day, two-way EHR sync for ~1M users

About

Backend engineer (3+ years) building production systems where reliability and AI meet: Python, Django, FastAPI, Celery, GraphQL, Kafka, AWS, MySQL, MongoDB, and LLM engineering (OpenAI GPT, Gemini / Vertex AI, LiteLLM). Backend SDE-1 at HealthifyMe in Bengaluru; previously SDE at Amazon. Highlights: • AI diet plans: designed a 3-layer LLM-grounding pipeline (macro recomputation, food-DB resolution, algebraic serving solver) that makes unreliable LLM output production-safe — ~900 plans/day to ~700 daily active users, each fully customized plan delivered in ~1 minute. • Migrated generation to an async FastAPI orchestrator: ~5s faster, zero-incident phased rollout (5/25/60/100%) verified by a 1e-9 golden-parity suite and shadow-run comparison. • Platform sync: built a loop-safe, idempotent two-way EHR sync (8 health-log domains over GraphQL + webhooks) that enabled ~1M users across two health-tech platforms, and designed UserLogEventRegistry, a pub-sub framework now reused by multiple services. • GLP-1 automation: replaced a daily manual ops task with a hands-off pipeline validating ~100 pen orders/day against doctor prescriptions — deterministic parser with an extraction-only LLM fallback, idempotent quota-safe third-party export. • At Amazon: designed and built an LWA (Login with Amazon) SDK onboarding pipeline in AWS CDK — the standardized path for Kepler app onboarding. I care about the unglamorous parts that make systems trustworthy: idempotency, backward compatibility, phased rollouts, and deterministic validation over LLM trust.

Location

Bengaluru

AI stack

anthropic apis

claude code

codex

grok

hermes agent

mcp

Work experience

Backend Software Engineer (SDE-1) at Healthify, Mar 2025 - Present · 1 y 5 mo, Bengaluru, Karnataka, India, Tech: Python, Django, Celery, FastAPI, GraphQL, MySQL, MongoDB, LiteLLM (OpenAI GPT-4o, Gemini / Vertex AI) • Designed a 3-layer LLM-grounding pipeline (macro recomputation, food-DB resolution, algebraic serving solver) guaranteeing nutritionally valid AI diet plans — ~900 plans/day to ~700 daily active users, delivered in ~1 minute. • Reduced diet-plan generation time by ~5s by moving I/O-bound LLM calls off Celery into an async FastAPI orchestrator, via a zero-incident phased rollout backed by a 1e-9 golden-parity suite. • Enabled ~1M users across two health-tech platforms by building a loop-safe, idempotent two-way EHR sync across 8 health-log domains over GraphQL and webhooks. • Designed UserLogEventRegistry, a reusable pub-sub framework that decouples user-log emission from consumption — adopted by Ria AI, coach highlights, and Healthie sync. • Automated daily GLP-1 order-vs-prescription validation (~100 orders/day) with a deterministic parser plus extraction-only LLM fallback and an idempotent, quota-safe Google Sheets export. • Fully owned the clinical charting-notes + AI-summary feature: webhook ingest, GraphQL fetch, MongoDB storage, LLM summaries, coach notifications, Ria AI context injection. • Reorganized the Coach-Copilot v2 API into 6 purpose-specific REST endpoints and built a provider-aware multimodal content pipeline (150+ MIME types, OpenAI + Vertex AI). • Owned the LLM model lifecycle: migrated deprecated Gemini/GPT models, upgraded the multi-provider LiteLLM integration, mitigated 429/rate-limit failures before outages.

Software Development Engineer at Amazon, Apr 2024 - Feb 2025 · 11 mos, Bengaluru, Karnataka, India, Tech: AWS, AWS CDK, C++, Python, S3, RDS • Designed and built the LWA (Login with Amazon) SDK onboarding pipeline in AWS CDK, making the SDK the standardized path for Kepler app onboarding. • Ran the onboarding pipeline at 99.9% uptime with 100% compliance with Kepler onboarding requirements for SDK externalization. • Integrated CDK with LivePipeline Template (LPT), enabling seamless, hands-off production deployments. • Implemented a comprehensive error-handling strategy across services — unhandled exceptions down ~95%, MTTR for critical issues down ~60%. • Worked across C++ and Python services on AWS infrastructure (S3, RDS), shipping all infrastructure as code with CDK.

Backend Development Engineer at Tailnode Technology, Sep 2023 - Mar 2024 · 7 mos, Remote, Tech: Node.js, TypeScript, Cassandra DB, Kafka, MySQL, Jenkins • Built the backend for the employee self-onboarding system (web + mobile), cutting manual onboarding effort by ~50% and improving time-to-productivity for new hires. • Resolved a critical MySQL race condition that produced duplicate employee codes by routing task processing through Kafka for sequential ordering. • Integrated secure OAuth2 Google Login for user authentication. • Hardened Node.js APIs with comprehensive exception handling, cutting production failures by ~30% within the first month. • Owned bug fixes and on-call support — rapid incident response and hotfix deployments maintaining ~90% service availability.

Software Development Engineer Intern at Amazon, Feb 2023 - Jun 2023 · 5 mos, Bengaluru, Karnataka, India, Tech: AWS (Redshift, EC2, S3, RDS, Beanstalk, CloudWatch), Python, Pandas, Glue, QuickSight • Built a cost-estimation tool for Kindle Device Metrics (KDM) using Python, Pandas, and AWS Glue that surfaced underutilized AWS resources — driving a ~20% reduction in internal AWS bills through decommissioning. • Collected and maintained AWS usage data across Redshift, EC2, S3, RDS, Beanstalk, and CloudWatch, and built a dashboard for querying detailed service and resource information. • Transformed large datasets into analysis-ready formats with Pandas, enabling resource cost analysis and optimization.

Python Development Intern at TechWhoop, Jan 2022 - Feb 2022 · 2 mos, Remote, Tech: Django, Python, GCP, Google Sheets API, Selenium, Bootstrap • Automated a manual Excel-based operations workflow end to end using Python/Django and the Google Sheets API on GCP, eliminating recurring manual ops effort and adding a daily run-results dashboard. • Built Django backends and Bootstrap front ends for three websites, integrating the Google Sheets API for result storage and the 2Captcha API over HTTP. • Kept crawler bots running reliably with rotating proxies to prevent bans.

Education

Bachelor of Technology - BTech, Computer Science, jaypee institute of information technology

Silverline Prestige School

Public links

github: https://github.com/hashharit