Available for new opportunities

Lalith Kunal
Bachu

AI/ML Engineer & Full-Stack Software Engineer

6 years building production-grade ML systems — from LoRA fine-tuning and LangGraph multi-agent orchestration to RAG pipelines and distributed Ray Serve inference. Passionate about shipping AI that works at scale.

View Projects Get In Touch GitHub LinkedIn
6
Years Experience
2
AI Platforms Built
40%
Hallucination Reduction
500+
Concurrent Requests

Technical Arsenal

Skills & Technologies

AI & ML Core
LLM Fine-Tuning LoRA / QLoRA RLHF RAG Pipelines Multi-Agent Orchestration LangGraph Agentic AI Hallucination Mitigation
Models & Frameworks
PyTorch TensorFlow HuggingFace LangChain scikit-learn Gemini 1.5 Flash Llama 4 Scout Stable Diffusion BiomedCLIP llama.cpp
Languages
Python TypeScript Java JavaScript SQL C++
Cloud & MLOps
AWS SageMaker Azure AZ-900 Docker Kubernetes Ray 2.9 MLflow W&B
Backend & Data
FastAPI Spring Boot Node.js PostgreSQL pgvector Supabase Redis Apache Kafka

Featured Work

AI & ML Projects

🏥
SynthoMed
AI-Powered Synthetic Medical Imaging Platform
↓ 40% Hallucinations
↓ 60% Manual Reviews
500+ Concurrent Req.
  • Full-stack platform generating clinically plausible synthetic medical cases via LoRA-fine-tuned Llama on 10K+ PubMed reports with a dual-source RAG pipeline.
  • LangGraph multi-agent system with 5 specialized agents: Medical Validator, Research (RAG), Case Generator, Imaging (Stable Diffusion), and Quality Auditor.
  • Self-correcting agentic loops — Quality Auditor routes failed cases back to the Generator with structured feedback, plus human-in-the-loop for edge cases.
  • Ray Serve on Kubernetes for distributed inference (zero cold starts) with Redis job queues; MLflow experiment tracking with BiomedCLIP & ChexPert metrics.
LangGraph LoRA RAG Stable Diffusion Ray Serve Kubernetes MLflow Redis BiomedCLIP
👗
AI Stylist
Multi-Modal Real-Time Personal Styling Engine
↑ 25% Style Accuracy
Sub-second Latency
12+ Concurrent LLM calls
  • LangGraph supervisor-pattern orchestrating 4 agents: Style Analyst (Gemini 1.5 multi-modal), Catalog Search (pgvector cosine similarity), Stylist, and Fit Validation.
  • Fine-tuned Llama 4 Scout via self-distillation using LoRA on 5K+ curated outfit pairings; W&B hyperparameter sweeps for full experiment traceability.
  • RLHF-ready telemetry pipeline capturing user selections (Save/Buy vs. Dismiss) as implicit preference signals for future DPO model alignment.
  • Sharded async LLM grading via ThreadPoolExecutor to batch 12+ garments into concurrent Groq API calls; zero-shot DistilBERT sentiment scores for trust signals.
LangGraph Llama 4 Scout LoRA Gemini 1.5 pgvector DistilBERT RLHF W&B Groq

Work History

Professional Experience

Software Engineer
New York State Department of Health
09/2022 – Present
Albany, NY
  • Led product scoping with clinical stakeholders, Product Owners, and engineering leads to define requirements for healthcare form management systems serving 10K+ users.
  • Designed RESTful APIs with Spring Boot and Spring Data in a microservices architecture for an Angular application managing healthcare patient forms.
  • Engineered automated CI/CD pipelines with AWS CodeBuild; integrated Apache Kafka for real-time data streaming and event-driven processing.
  • Implemented OAuth 2.0 authentication with Spring Security; built automated SMTP notification services, reducing manual follow-up overhead by 30%.
Software Engineer
Mastercard
04/2021 – 08/2022
New York, NY
  • Built a high-throughput API Gateway on NGINX reverse proxy with Lua-nginx module, implementing load balancing algorithms and SSL termination.
  • Improved REST API performance by 100% through TCP connection tuning and CDN optimization, securing traffic with JWT and API key authentication on OAuth 2.0.
  • Managed Docker-based development environments with Jenkins CI/CD pipelines; utilized Splunk for log analysis across gateway infrastructure.
Java Developer
Modivcare
10/2020 – 03/2021
Denver, CO
  • Developed a scalable, queue-based document processing service using JMS with AWS S3 integration for microservices file uploads and async job processing.
  • Built enterprise-grade applications using the J2EE stack (Spring, Hibernate, Spring JDBC) with a DB2 database.

Academic Background

Education & Certifications

M.S. Computer Science & Data Science
University of Cincinnati, OH
08/2019 – 04/2021 GPA 3.8
B.Tech. Information Technology
BV Raju Institute of Technology (BVRIT), Hyderabad
05/2015 – 04/2019 GPA 3.6
☁️
Microsoft Azure Fundamentals
AZ-900 · Microsoft

Let's Connect

Get In Touch

Open to AI/ML engineering roles, research collaborations, and interesting problems. Reach out — I'd love to chat.

kunallalith0110@gmail.com (513) 884-1282 github.com/Kunal0110 linkedin/lalithkunal-b