Sai Kiran Vepamani
Lead Engineer - Gen AI
Lead Engineer - Gen AI at Bosch Global Software Technologies with 6+ years of experience spanning Generative AI, LLM fine-tuning, functional safety (FuSA) AI, and cloud-native platform development. Led cross-functional teams building AI-powered products including RAG platforms, HARA AI for safety-critical automotive systems, voice-enabled developer tools, and multi-tenant SaaS infrastructure. Patent co-author, hackathon winner (Bosch AWS 2025, AppsForBharat 2025), and Bharat Mobility Expo 2025 presenter.
Professional Experience
Lead Engineer - Gen AI at Bosch Global Software Technologies
Mar. 2023 - Current — Bangalore, India
- Won Bosch AWS Hackathon 2025 for "Project IQ," an AI-powered training & knowledge management system with 90% accuracy.
- Won AppsForBharat 2025 hackathon for developing impactful solutions for India.
- Co-authored patent on detunneling data from ethernet frames (Patent No: 202541072960).
- Represented Bosch at Bharat Mobility Expo 2025, showcasing a Vehicle Assistant AI for SDV apps.
- Leading a cross-functional team of 8+ engineers across Gen AI, automotive middleware, and cloud platform initiatives.
- Built HARA AI, a safety-relevant AI system for Hazard Analysis and Risk Assessment per ISO 26262.
- Architected BRICK, an advanced document analysis platform featuring PathRAG and MinerU for high-precision retrieval.
- End-to-end architected HireStream, a recruitment OS using zero-shot learning, reducing manual screening by 80%.
- Performed Parameter-Efficient Fine-Tuning (PEFT) on Qwen 7B+ models using QLoRA/LoRA with 4-bit quantization.
- Developed SDX Assistant, a voice-enabled developer tool with STT/TTS and multi-turn conversation.
- Built SDX-A2L Signal Monitor with Two-Stage AI Search (20x faster) for parsing 100MB+ A2L files.
- Architected a Serverless Multi-Tenant SaaS platform using AWS CDK, reducing onboarding time by 90%.
SDE - II at HashedIn by Deloitte
Aug. 2021 - Mar. 2023 — Bangalore, India
- Built cloud-native content normalization platform for Thomson Reuters legal domain using AWS serverless architecture.
- Developed serverless applications using AWS Lambda, API Gateway, RDS, DynamoDB, and CloudFormation.
- Developed annotation component in Angular that reduced document processing time from 7 days to 4 hours.
- Architected AWS serverless cloud infrastructure with relational and non-relational database integration.
- Collaborated with US legal domain stakeholders for requirement gathering and defining technical user stories.
Software Engineer at Zapcom Solutions
Feb. 2020 - July 2021 — Bangalore, India
- Built REST APIs using Python/Django DRF and migrated database from Kinvey to Django models.
- Implemented FCM push notifications and Celery-based async task processing.
- Built Content Management service handling 1000+ client devices in WSO2.
- Redesigned 70% of UI improving user experience.
- Wrote automation test scripts reducing manual testing effort by 35%.
Full Stack Intern at Zapcom Solutions
Jan. 2019 - Jan. 2020 — Bangalore, India
- Redesigned Zapcom portal with responsive landing pages.
- Built scene classification model using Places365 dataset.
- Developed web scraping pipeline extracting 1000+ reviews from 540+ client hotels.
- Integrated Facebook Developer API for automated review extraction.
Key Projects
HARA AI
Safety-relevant AI system for Hazard Analysis and Risk Assessment that automates identification of safety goals and ASIL classifications per ISO 26262.
Technologies: Python, LangChain, RAG, ISO 26262, FastAPI
BRICK
Advanced document analysis platform featuring PathRAG and MinerU for high-precision retrieval from unstructured PDFs with structural preservation.
Technologies: PathRAG, MinerU, Python, React, FastAPI
HireStream
Recruitment OS matching associate skills against JDs using zero-shot learning, reducing manual screening by 80%.
Technologies: Zero-Shot Learning, Python, React, PostgreSQL
SDX Assistant
Voice-enabled developer tool with STT/TTS and multi-turn conversation for natural language code generation in automotive context.
Technologies: STT/TTS, LLM, Python, WebSocket, VSS
JobsChange.com
AI-powered career platform that tailors resumes to job descriptions, generates cover letters, and provides mock interview preparation.
Technologies: Next.js 15, React 19, TypeScript, Firebase, OpenAI
MakeDemos.com
Desktop screen recording application for creating polished product demos with auto-zoom, 50+ visual effects, and 4K export.
Technologies: Electron, React, PixiJS, Web Codecs API, Zustand
Technical Skills
Languages
Python, Java, C++/C, Rust, TypeScript, Node.js, MySQL, PostgreSQL, Bash
AI/GenAI
LLM Fine-Tuning (QLoRA/LoRA/PEFT/RLHF), RAG/PathRAG, Prompt Engineering, Agentic AI, DSPy, Agno, LangChain, LangSmith, LangGraph, Hugging Face, Vector DBs (FAISS/ChromaDB)
ML/Data
PyTorch, Tensorflow, Transformers, Scikit-learn, OpenCV, Pandas, Numpy, Optuna, MLflow, MLOps
Frameworks
FastAPI, Django, Flask, Spring Boot, Angular 19, Next.js, React, Flutter, Socket.IO, SQLAlchemy
Cloud/DevOps
AWS (Lambda/CDK/S3/Cognito), Docker, Kubernetes, Helm, Jenkins, Git, CI/CD, SonarQube, Gradle
Protocols/Tools
MCP, A2A, gRPC, Protobuf, MQTT/mTLS, JWT, REST, WebSocket, VSS, CAN/DBC, Playwright, Postman, Jira
Achievements
Patent Co-Author (2025): "A control unit for detunneling of data from an ethernet frame" (No: 202541072960)
Bosch AWS Hackathon Winner (2025): "Project IQ" - AI-powered training & knowledge management system
AppsForBharat Winner (2025): Developing impactful solutions for India
Bharat Mobility Expo Presenter (2025): Showcased Vehicle Assistant AI for SDV apps on HMI and HPCs
Education
B.Tech in Computer Science and Engineering — JNTUA (Jawaharlal Nehru Technological University Anantapuramu) (Aug. 2015 - May 2019), 7.74 GPA