EnpalBusiness Analyst · clustered customer needs and workflow pain points
AMON SISOWATH / PORTFOLIO 2026
I build AI systems end to end.
From information-retrieval research to production AI architectures, rapid prototypes, and teams that ship. This is the full path behind my work.
RESEARCH24.6M-passage retrieval index
BUILDINGHack-Nation Track Winner
LEADERSHIPLed eight engineers at TUM.ai
QUERY / ENTITY RETRIEVALTOP-K: 100
How does entity popularity affect retrieval?
01
Popular entity
0.92
02
Known entity
0.76
03
Mid-tail entity
0.44
04
Long-tail entity
0.18
24.6Mpassages
06benchmarks
BM25 + FAISSretrievers
01
Information Retrieval · Current
Measuring who gets retrieved.
At UC Berkeley Sky Computing Lab, I research popularity bias and long-tail failure in retrieval-augmented generation.
QUESTIONHow does entity popularity shape retrieval and generation?
SCALE24.6M Wikipedia passages across six benchmarks
RETRIEVALBM25/BM25+, FAISS, sparse + dense retrieval, MRR, recall
MODELINGPyTorch fusion classifier built on a LoRA-fine-tuned BERT encoder
PRODUCTION SERVER ARCHITECTURE / ARTICLYRUNNING
01Publisher feeds
02Topic ranking
03Dense retrieval
04Script generation
05Editorial approval
06Podcast output
MCP ORCHESTRATIONMULTI-CUSTOMER CONFIGHUMAN IN THE LOOP
02
Work Experience · Most Recent
Engineering across the stack.
Most recently at ARTICLY, I independently redesigned the production server architecture and AI processing pipeline behind an automated podcast platform for newspaper publishers.
OWNERSHIPSolo server architecture redesign and implementation
PIPELINETopic selection → retrieval → scripts → approval → audio
ARCHITECTUREMCP orchestration, dense similarity ranking, multi-tenant configuration
MOST RECENT / MARKET LINK NETWORK1,000+ MARKETS
EMBEDDING CLUSTER 04ρ 0.87ARBITRAGE SIGNAL03
Hackathons · Rapid Building
Building useful systems under pressure.
Most recently, in 39 hours, our team built Market Link Network to surface hidden relationships across more than 1,000 prediction markets.
TOP 10of 300+ global teamsTRACK WINNER
COREOpenAI embeddings + HDBSCAN clustering
STACKFastAPI, Supabase, Google Cloud Run, D3.js
REASONINGGemini Flash with sub-two-second responses
LEADERSHIP SYSTEMTEAM → COMMUNITY
08ENGINEERS
PLATFORM
RAG ASSISTANT
TUM.AI→100+ USERS+PUBLIC INSTITUTIONS
04
Leadership + Institution Building
From public life to technical teams.
Before leading engineers, I learned to build institutions: creating political access for students, serving locally, and taking responsibility for communities.
TUM.AILed eight developers building a platform and RAG assistant for 100+ users
FOUNDEDCo-founded Liberale Schüler Berlin
SERVEDJunge Liberale district board · Tafel volunteer
RECOGNITIONFriedrich Naumann Scholar · selected among the top 3%
ENTREPRENEURSHIP LOOP2022 - PRESENT
Young Founders Network
Innovis VC
muffin · ARTICLY
Products · Hackathons
1st / 80BCG B@S × Waterdrop consulting competition
05
Entrepreneurship
Learning how ideas become leverage.
The Young Founders Network introduced me to builders creating impact beyond their immediate reach. Venture analysis and startup roles taught me how those ideas are evaluated and operated.
YFNMember since 2022 · connected to founders and the startup ecosystem
INNOVIS VCEvaluated deep-tech startups for a database used by investors including Earlybird and Pictus
OPERATINGWorked inside startups at muffin and ARTICLY
PRODUCTDeveloped commercial judgment at CHECK24 and Enpal
AI LEARNING PATH3 INSTITUTIONS
BERUC Berkeley
- Agentic AI
- Deep Learning
- Leadership
SINNational University of Singapore
- Machine Learning
- Feature Engineering
- Information Security
MUCTechnical University of Munich
- Statistics
- Machine Learning Foundations
- LLM Fine-Tuning
- Linear Algebra
06
AI Education · Three Institutions
Learning AI across three ecosystems.
TUM established my statistics, machine-learning, and mathematical foundations; NUS deepened my applied ML practice; and Berkeley connected implementation with advanced AI systems and research.
BERKELEYAgentic AI · Deep Learning · Leadership
NUSMachine Learning · Feature Engineering · Information Security
TUMStatistics · Machine Learning Foundations · LLM Fine-Tuning · Linear Algebra
07 / DETAILED VIEWBuilt across disciplines.
Built across disciplines.
Focused on Applied AI.
Every role, initiative, and institution behind the six chapters.
Engineering + Research
ARTICLY
AI Engineer · production server architecture overhaulUC Berkeley Sky Lab
Independent Researcher · information retrievalTUM.ai
Development Team Lead · eight-person teamSAP
Junior Software Engineer InternProduct + Entrepreneurship
Enpal
Business Analyst InternInnovis VC
Associate · deep-tech startup analysisCHECK24
Product Managermuffin
Junior Product Manager InternYoung Founders Network
MemberLeadership + Public Life
Liberale Schüler Berlin
Co-founderJunge Liberale
District Board MemberJunge Liberale
MemberTafel Deutschland
VolunteerFriedrich Naumann Foundation
Scholar · top 3%Education
UC Berkeley
Agentic AI · Deep Learning · LeadershipNational University of Singapore
Machine Learning · Feature Engineering · Information SecurityTechnical University of Munich
Statistics · Machine Learning Foundations · LLM Fine-Tuning · Linear AlgebraPYTHONPYTORCHBERTLORARAGMCPFAISSBM25+FASTAPINEXT.JSGCP