Jaswanth Tikkireddy, home

Changelog

Career history, written as release notes: what each role shipped, changed, fixed and contributed to. Most recent first.

v1.0Software Engineer

Bettrlabs

Jan 2026 - present
Chennai, on-site

AI and backend engineering at Bettrlabs: legal and regulatory RAG, compliance AI, hybrid search, payments and the core backend platform.

retrieval accuracy on FSSAI documents
95%

added8

  • Legal and regulatory RAG, designed and built independently: automated ingestion, context-enriched chunking, hybrid retrieval, reranking and 14+ weighted embedding fields per chunk, reaching 95% retrieval accuracy on FSSAI documents.
  • Model-aware context selection and automated document versioning, run as an agent workflow with human approval on high-impact stages.
  • The core architecture of the PLM compliance engine: regulatory RAG, structured validation, compliance reporting, a multi-LLM council and a knowledge-graph and data layer.
  • Formulation search retrieval architecture, built independently for 10K+ formulations: database-first lookup with trigram matching, 11+ dense embedding fields per chunk, BM25, reciprocal rank fusion, reranking and an agentic fallback.
  • The Razorpay payment backend, built independently and end to end: verification, webhooks, reconciliation, idempotent settlement and recovery from delayed or inconsistent states.
  • Major parts of the async FastAPI backend: a layered codebase with one API response contract, PostgreSQL with Alembic migrations, a LiteLLM gateway with model fallbacks that logs every call’s tokens and cost, and heartbeat-lease background jobs that survive restarts.
  • Unit, API and end-to-end test coverage for the backend.
  • The agent harness loop: execution flow, tool selection and skill retrieval.

changed1

  • Built the PLM backend and integrated the compliance architecture into the production platform.

shipped

v0.9Software Engineer Intern

Bettrlabs

Oct 2025 - Dec 2025
Chennai

RAG, backend and full-stack engineering on the platform’s early AI workflows and production infrastructure, leading into the full-time role.

added4

  • The initial RAG pipelines for document ingestion, embeddings and retrieval, which became the foundation for later retrieval work.
  • LLM token-usage tracking and structured logging for production debugging.
  • Full-stack features with FastAPI, PostgreSQL and SQLite, React and Next.js: API routing, database flows and data-driven UI.
  • Backend deployment support with FileZilla, PuTTY and SSL setup.

v0.1AI/ML Remote Intern

INTRAIN TECH

Sep 2023 - Nov 2023
Remote

better prediction accuracy
20%
user ratings explored
1M
clean records
383K
better decision-making accuracy
15%

added4

  • A hybrid book recommendation system using collaborative filtering, content-based filtering and SVD, improving prediction accuracy by 20%.
  • Exploratory analysis of 1 million user ratings, surfacing demographic and rating-behaviour patterns.
  • A clean 383K-record dataset, after resolving 40% missing values, removing outliers and merging sources.
  • Power BI dashboards on demographics, age distributions and top books, improving decision-making accuracy by 15%.

rootCSE (AI & ML)

VIT-AP University

education